thapa sunil

DETECTION AND MAPPING OF INCIDENCE OF VISCUM ALBUM IN PINUS SYLVESTRIS FOREST OF SOUTHERN FRENCH ALPE USING SATELLITE AND AIRBORNE OPTICAL IMAGERY

THAPA, SUNIL February, 2013

SUPERVISORS: DR. HEIN A.M.J. VAN GILS DR. YOUSIF HUSSIN

Thesis submitted to the Faculty of Geo-Information Science and Earth Observation of the University of Twente in partial fulfilment of the requirements for the degree of Master of Science in Geo-information Science and Earth Observation. Specialization: Natural Resource Management SUPERVISORS: Dr. Hein A.M.J. van Gils Dr. Yousif Hussin THESIS ASSESSMENT BOARD: Prof. Dr. Andrew K. Skidmore (Chair) Dr. J.F. Duivenvoorden (External Examiner, Institute for Biodiversity and Ecosystem Dynamics (IBED) – University Amsterdam)

DETECTION AND MAPPING OF INCIDENCE OF VISCUM ALBUM IN PINUS SYLVESTRIS FOREST OF SOUTHERN FRENCH ALPE USING SATELLITE AND AIRBORNE OPTICAL IMAGERY

THAPA, SUNIL Enschede, the Netherlands, February, 2013

DISCLAIMER This document describes work undertaken as part of a programme of study at the Faculty of Geo-Information Science and Earth Observation of the University of Twente. All views and opinions expressed therein remain the sole responsibility of the author, and do not necessarily represent those of the Faculty.

i

ABSTRACT

Increasing incidence rate of Viscum album in Pinus sylvestris is one of the major problems in the conservation and sustainability of Pinus sylvestris forest in the Alps. Mostly plantated forests are highly susceptible to biological invasion as compared to natural forests. The study area covered with coniferous plantation (mostly Pinus sylvestris in lower elevation), a part of Bois noir in the South Western French Alps is highly affected by semi-parasites Viscum album. Consequences like swelling of the branch, bending of tree structure; tree mortality of Pinus sylvestris is in alarming rate in the study area due to Viscum album incidence. For the management and minimization of the biological invasion, detection and mapping plays a key role in the forest conservation. Detection and mapping of a biological invasion through remote sensing is a challenge to researchers to overcome these issues. Advancement of very high resolution (VHR) satellite imagery and aerial imagery with application of remote sensing and GIS technologies has shown a promising result in the detection, mapping and monitoring of the forest health. In this research digital aerial Ortho imagery (15cm resolution) and VHR satellite imagery WorldView-2 (panchromatic 0.5m and multispectral 2m) was used to detect and map the presence of Viscum album in the Pinus sylvestris forest by pixel based maximum likelihood classifier. Distribution of Pinus sylvestris forest was successfully mapped with higher accuracy (96%) and kappa coefficient of 0.84 on WorldView-2 optical imagery. Pinus sylvestris in presence of Viscum album have low spectral reflectance in all bands however NIR1, NIR2 and red edge of WorldView-2 have the higher ability to discriminate the Viscum album. Similarly, the vegetation index NDVI 85 (band combination of red and NIR2) has the potential to discriminate the Viscum album. Furthermore, the results indicated that the Viscum album has a negative correlation and significant relationship (r=-0.5135; p<0.01) with the elevation whereas a significant positive relationship (r=0.52; p<0.01) with DBH of Pinus sylvestris. Weak but statistically significant multiple regression and a logistic regression was developed by using elevation and DBH to model the incidence of Viscum album in Pinus sylvestris trees. Overall classification accuracy (86%) and kappa coefficient (0.52) was achieved in WorldView-2 image by applying pixel based maximum likelihood algorithm for the detection of Viscum album in Pinus sylvestris forest. A comparison of the 2m resolution WV-2 and 0.15cm resolution orthophoto classification outputs shows that the WV-2 imagery of lower spatial but higher spectral resolution gave a moderate higher (86%) classification accuracy. The study reveals the high potential of high resolution optical imagery for detection and mapping tree infestation. Detection and mapping of such a biological invasion serves good information for the better management of the forest. Keywords: Detection and mapping, Pinus sylvestris, Viscum album, Optical imagery, Biological invasion

ii

ACKNOWLEDGEMENTS

This study has been accomplished with the encouragement, cooperation, guidance and suggestions received from various esteemed organization and distinguished personalities, both moral and practical support. I acknowledge this help with gratitude and humbly regret that the individuals are too many to mention here. I am deeply honoured and would like to express my gratitude to the EU Erasmus Mundus Mobility for Asia (EMMA Lot 11) for providing me the fellowship to pursue MSc Degree in the Faculty of Geo-information Science and Earth Observation (ITC), the University of Twente. I am highly in debt and grateful to Dr. H.A.M.J. Hein van Gils, my first supervisor, for introducing me the research topic and for his continuous encouragement, patient, support and invaluable guidance throughout the research. I would like to extend my sincere appreciation and gratitude to my second supervisor Dr. Yousif Hussin for his continuous support, valuable suggestions and assistance during my whole research. Both of my supervisors have broadened and deepened my knowledge and introduced me in the scientific world. I wish to extend my profound appreciation to Dr. Michael Weir, Course Director, NRM for his valuable support, guidance, co-operation and facilitating my research. I would like to express my deep gratitude to Prof. Dr. Andrew Skidmore, his advice and a suggestion has been a great help in formulating the research in a scientific way. I would also like to offer my sincere thanks to Dr. Chudamani Joshi, Drs. E.H. (Henk) Kloosterman and Mr. Vinod Kumar for their immense support and guidance and constructive suggestions during the development of this research proposal. I am particularly indebted to Mr. Grigorijs Goldbergs and Mr. Collins Kukunda for their immense help in Ortho-rectification of the aerial photo and geometric correction of WorldView-2 imagery. My sincere thanks to Dr. Thomas Groen, and Mr. Chandra Prasad Ghimire for their guidance on statistical analysis of the research. I would like to express my great appreciation to Lina Maria Estupinan Suarez, Collins Kukunda and Anahita Khosravipour for their full support, valuable suggestions, field guidance, moral support and technical assistance (ERDAS and GIS) throughout the research. I am truly thankful to Mr. Pradeep Sapkota Upadhyaya and for his technical support on GIS application for this research. I am equally thanks to Marnes Rasel, Arun Poudyal, , Dhan Shrestha, Anuj Pradhan, Shrota Shrestha, Joana pinto, Joaquin Duque, Yogendra Karna, Geliah Gloria, Metadel, Efthymia Pavlidou, Maryam Gol, Sabah Sabaghy, Manuel Garcia, Bayarma Enkhtur, Islam Fadel, , Hari Dhonju, Sunita and Reshma for their support and suggestions which helps me to upgrade my research. A word of appreciation also goes to all NRM and GEM classmates for their continuous support and sharing their cheerful moments. Special thanks to Nepali society for their warm regards, support and encouragement during my study in ITC. I express my deepest appreciation to my beloved Dutch friends Mattheus Francinus Visser, Petra Sloot, Christien Warners, Wilma Dierx, Netty Kuijken members of Nassaukerk Church Amsterdam, for their continuous moral support, kindness and warm hospitality throughout my stay in The Netherlands. Last but not the least, deepest appreciation to my respected parents and sisters who have been a source of inspiration to every piece of my work. Sunil Thapa (February, 2013) Enschede, the Netherlands

iii

...IN LOVING MEMORY OF MY LATE GRANDMOTHER SARASWOTI THAPA

iv

TABLE OF CONTENTS

ABSTRACT ................................................................................................................................................................................i ACKNOWLEDGEMENTS ................................................................................................................................................ii

LIST OF FIGURES .........................................................................................................................................................vi LIST OF TABLES...........................................................................................................................................................vii LIST OF APEENDICES ..............................................................................................................................................viii 1. INTRODUCTION .......................................................................................................................................... 1

1.1. Pinus sylvestris (Scots pine)..............................................................................................................................1 1.2. Viscum album (Mistletoe) ................................................................................................................................1 1.3. Overview of Viscum album impact on plant species .......................................................................................2 1.4. Viscum album impact on Pinus sylvestris ........................................................................................................3 1.5. Application of remote sensing for tree species classification and pest detection in forestry.........................3

1.5.1. Use of spectral reflectance and vegetation indices in monitoring plant health.........................................4 1.6. Problem statement and justi fication ...............................................................................................................5 1.7. Research objectives :.........................................................................................................................................7

1.7.1. Specific objectives: ......................................................................................................................................7 1.8. Research questions ..........................................................................................................................................7 1.9. Hypothesis: .......................................................................................................................................................7 1.10. Research innovation .........................................................................................................................................7

2. DESCRIPTION OF STUDY AREA ......................................................................................................... 8 2.1. Background.......................................................................................................................................................8 2.2. Study Area ........................................................................................................................................................8

2.2.1. Climate.........................................................................................................................................................8 2.2.2. Geology:.......................................................................................................................................................8 2.2.3. Vegetation ...................................................................................................................................................9

2.2.3.1. Pinus sylvestris L. (Le Pin sylvestre):.....................................................................................................9 2.2.3.2. Mountain pine [Pinus uncinata Mill. ex Mirb. (Le Pin a crochet)]: ....................................................10 2.2.3.3. Larix decidua Miller.: ..........................................................................................................................10 2.2.3.4. Picea abies L. (Karsten):......................................................................................................................10

3. MATERIALS AND METHODS .............................................................................................................. 11 3.1. Satelli te data...................................................................................................................................................11 3.2. Digi tal aerial Ortho image and LiDAR data.....................................................................................................11 3.3. Field material..................................................................................................................................................11 3.4. Processing Software .......................................................................................................................................11 3.5. Methods .........................................................................................................................................................12

3.5.1. Sampling Design ........................................................................................................................................12 3.5.2. Field techniques : .......................................................................................................................................13

3.6. Image pre-processing .....................................................................................................................................15 3.6.1. Geometric correction ................................................................................................................................15 3.6.2. Image fusion and enhancement................................................................................................................15

3.7. Fieldwork data analysis ..................................................................................................................................15 3.7.1. Vegetation analysis....................................................................................................................................15 3.7.2. Tree diversi ty analysis: ..............................................................................................................................16 3.7.3. Statis tical analysis......................................................................................................................................16 3.7.4. Image classification ...................................................................................................................................17

3.7.4.1. Supervised pixel based maximum likelihood classification (MLC):....................................................17

v

3.7.5. Extraction of canopy spectra of Pinus sylvestris in presence and absence of Viscum album from WV-2 image: 17 3.7.6. Computing and analysis of spectral vegetation indices............................................................................ 17 3.7.7. Supervised pixel based maximum likelihood classification (MLC) of WV-2 and Ortho image:................ 18

3.8. Accuracy assessment: .................................................................................................................................... 18 4. RESULTS.......................................................................................................................................................... 20

4.1. Tree species in the forest............................................................................................................................... 20 4.2. Tree diversi ty:................................................................................................................................................. 20 4.3. Descriptive s tatisics of Pinus sylvestris forest................................................................................................ 21

4.3.1. Relationship of DBH and height ................................................................................................................ 21 4.4. Descriptive anlysis of presence of Viscum album in Pinus sylvestris ............................................................ 22 4.5. Relationship of Viscum album in Pinus sylvestris with elevation and DBH ................................................... 23 4.6. Multiple regression ........................................................................................................................................ 25 4.7. Logis tic regression.......................................................................................................................................... 25 4.8. Image processing ........................................................................................................................................... 26 4.9. Image classification ........................................................................................................................................ 26 4.10. Spectral analysis of Pinus sylvestris in the presence and absence of Viscum album .................................... 28 4.11. Analysis of vegetation indices (NDVIs)........................................................................................................... 28 4.12. Classification of presence and absence of Viscum album in Pinus sylvestris in the WV-2 image ................. 29 4.13. Classification of presence and absence of Viscum album in Pinus sylvestris on ai rborne multispectral Ortho image 31

5. DISCUSSION.................................................................................................................................................. 33 5.1. Pinus sylvestris plantation forest ................................................................................................................... 33 5.2. Relationship of Viscum album in Pinus sylvestris with elevation and DBH ................................................... 33 5.3. Analysis of spectral reflectance and vegetation indices................................................................................ 34 5.4. Image classification and accuracy assessment .............................................................................................. 34

5.4.1. Pinus sylvestris distribution map .............................................................................................................. 34 5.4.2. Detection of Viscum album in Pinus sylvestris in WorldView-2 and the Ortho image............................. 35

5.5. Sources of error in image classification for detection of Viscum album in optical imagery ......................... 35 6. CONCLUSIONS AND RECOMMENDATIONS............................................................................. 37

6.1. Conclusions .................................................................................................................................................... 37 6.2. Recommendations ......................................................................................................................................... 37

LIST OF REFERENCES .............................................................................................................................................. 39 LIST OF APPENDICES: .............................................................................................................................................. 45

v i

LIST OF FIGURES Figure 1: Distribution of (A) Pinus sylvestris and (B) Viscum album in Europe.......................................................... 3 Figure 2: Physical structure of cones of Pinus sylvestris, Pinus uncinata and Pinus mugo.......................................... 10 Figure 3: Flowchart of research ......................................................................................................................................... 12 Figure 4: Field sample points locations in Ortho photo............................................................................................. 14 Figure 5: Frequency of tree species in sampled plot .................................................................................................... 20 Figure 6: Density of tree species in sampled plot ......................................................................................................... 20 Figure 7: Pinus sylvestris tree distribution parameters (A) DBH, (B) Elevation, (C) Height ........................... 21 Figure 8: Pinus sylvestris tree parameters (A) DBH, (B) Elevation, (C) Height .................................................. 21 Figure 9: Scatter plot showing relationship of DBH and height of Pinus sylvestris ............................................... 21 Figure 10: Relative frequency of Pinus sylvestris having different density level of Viscum album........................ 22 Figure 11: Incidence of Viscum album in Pinus sylvestris trees....................................................................................... 22 Figure 12: Density of Pinus sylvestris having different level of Viscum album .......................................................... 22 Figure 13: Basal area and of Pinus sylvestris having different level of Viscum album .............................................. 22 Figure 14: Relative frequency and relative abundance of Pinus sylvestris having different density class of Viscum album ............................................................................................................................................................................ 23 Figure 15: Incidence of Viscum album and elevation (m) (plots) ............................................................................... 23 Figure 16: Incidence of Viscum album in different elevation ranges (plots) ........................................................... 23 Figure 17: Presence of Viscum album in Pinus sylvestris per DBH class (plots) ....................................................... 24 Figure 18: Boxplot of Pinus sylvestris DBH of presence/absence of Viscum album (plots) ................................. 24 Figure 19: Percentage of occurrence of different density level of Viscum album on different DBH classes of Pinus sylvestris ............................................................................................................................................................................ 24 Figure 20: Image location showing 'before' and 'after' geometric correction ....................................................... 26 Figure 21: Distribution map of Pinus sylvestris in Bois noir, Barcelonnette France (WV-2 image) .................. 27 Figure 22: Spectral reflectance of Pinus sylvestris in the presence and absence of Viscum album ....................... 28 Figure 23: Boxplot of reflectance value of three different vegetation indices for presence and absence of Viscum album in Pinus sylvestris ............................................................................................................................................. 29 Figure 24: Distribution map (WV-2) of presence of Viscum album in Pinus sylvestris forest ............................... 30 Figure 25: Distribution map (Ortho image) of presence of Viscum album in Pinus sylvestris forest .................. 32

v ii

LIST OF TABLES List of equipment ............................................................................................................................................ 11 Table 1. List of software and its purpose of usage ................................................................................................. 11 Table 2. NDVI Vegetation Indices............................................................................................................................. 18 Table 3. T-test for determining relationship of presence of Viscum album with elevation and DBH ....... 25 Table 4. ANOVA for determining relationship of presence of Viscum album with elevation and DBH. 25 Table 5. Summary of multiple regression for prediction of Viscum album in Pinus sylvestris trees .............. 25 Table 6. Summary of goodness of fit ......................................................................................................................... 25 Table 7. AIC value of models....................................................................................................................................... 25 Table 8. Coefficients and significances of predictors ............................................................................................ 26 Table 9.

Species classification accuracy assessment ............................................................................................. 27 Table 10. Reflectance value of three vegetation indices........................................................................................ 28 Table 11. Accuracy assessment for the WV-2 image for presence and absence of Viscum album in Pinus Table 12.

sylvestris ...................................................................................................................................................................... 29 Kappa value for presence and absence of Viscum album for classified WV-2 image .................. 29 Table 13. Accuracy assessment for Ortho image for presence and absence of Viscum album in Pinus Table 14.

sylvestris ...................................................................................................................................................................... 31 Kappa value for presence and absence of Viscum album for classified Ortho image.................. 31 Table 15.

v iii

LIST OF APEENDICES Appendix 1. Taxonomy of Pinus sylvestris and Viscum album : .............................................................................. 45 Appendix 2. Life cycle of Viscum album ...................................................................................................................... 45 Appendix 3. Location map of the study area ............................................................................................................ 46 Appendix 4. Features of WV-2 image......................................................................................................................... 47 Appendix 5. Field data sheet ........................................................................................................................................ 48 Appendix 6. Spatial location of sample plots ............................................................................................................ 49 Appendix 7. LiDAR derived tree Canopy Height Model (CHM) of sample plot 30................................ 50 Appendix 8. Descriptive statistics of the Bois noir ................................................................................................. 50 Appendix 9. Histogram of Pinus sylvestris tree parameters...................................................................................... 51 Appendix 10. Descriptive statistics of Pinus sylvestris tree parameter................................................................ 52 Appendix 11. DBH class of Pinus sylvestris............................................................................................................... 52 Appendix 12. Descriptive statistics of presence and absence of Viscum album in Pinus sylvestris .............. 52 Appendix 13. Kappa statistics of classified species ............................................................................................. 52 Appendix 14. Field pictures ........................................................................................................................................ 53

DETECTION AND MAPPING OF INCIDENCE OF VISCUM ALBUM IN PINUS SYLVESTRIS FOREST IN SOUTHERN FRENCH ALPE USING SATELLITE AND AIRBORNE OPTICAL IMAGERY

1

1. INTRODUCTION 1.1. Pinus sylvestris (Scots pine)

Pinus sylvestris (L.) commonly known as Scots pine or Scotch pine, a highly diverse species (Xenakis et al., 2012) found from sea level to 2440m and ranges from Montane forests to Steppe (Cregg & Zhang, 2001). It’s distribution extends from 2700km in north to south and 14,000km in east to west across Asia and Europe (Roche et al., 2009) and covers 28 million hectares of forests in Europe (Vacchiano et al., 2008). Pinus sylvestris can grow up to 36m high and can have wide range of mature growth form (Trees for Life, 2012). It is a light demanding and a stress tolerant species (Richardson, 1998), prefers sandy and loamy soils and can survive long in drought conditions. It can live nearly 1,000 years and 200 to 400 years of ages are normal (Jones, 1945). However its growth rates are reduced in alkaline soil and seedling growth is optimum in acidic soil (Carter 1987). Pinus sylvestris plays a vital role in the diversity of the forest and is one of the major plant species on which many other species depend. Pinus sylvestris holds significant aesthetic natural, medicinal, timber, commercial and ornamental values. It is a keystone species and has relationships with various flora and fauna (Trees for Life, 2012). Pinus sylvestris has a direct and indirect role of host species for many plants, insects, birds and animals. For instance Scottish crossbill (Loxia scotica), endemic bird of UK is confined to the pinewoods (Trees for Life, 2012). It also plays a significant role in controlling rocks slides, landslides, erosion, windbreaks (Dobbertin et al., 2005a; Vacchiano et al., 2008).

1.2. Viscum album (Mistletoe) Viscum album austriacum commonly known as mistletoe is a “mostly globose perennial evergreen shrub with persistent haustoria in the host” (Zuber, 2004b). Viscum album has a wide range of distribution in Asia and Europe (Hawksworth & Scharpf, 1986) “extends from 10° W to 80 E ad from about 60° N (max 59° 38’ N) to 35° S” Zuber (2004a). There are four subspecies of Viscum album in Europe with varied host species (Zuber, 2008) . Furthermore, Barney et al. (1998) pointed that Viscum album have at least 452 plant species as its host species. Viscum album fully depends on their host tree for water and nutrients. They are hemi-parasites (Watson, 2001) for which the transpiration rates and stomatal conductance is higher whereas leaf water potential is lower than host species. However, Viscum album has its own functional chlorophyll (Tsopelas et al., 2004) and partially assimilate their own carbon by photosynthesis (Zweifel et al., 2012). Normal years for starting of sexual reproduction is at five years and

Photo 2: Fruit of Viscum album (courtesy: Dr. Hein van Gils)

Photo 1: Pinus sylvestris forest


2

March to April (Wallden, 1961; Zuber, 2008) is the main seasonal time for the flowering of Viscum album. Host habitat, size and canopy characteristics govern the incidence of Viscum album and it can grow at an average temperature above 15°c to -8°c in warmest and coldest month respectively (Dawson et al., 1990; Zuber, 2004b). Wind and birds are the main key agents for its pollination and dispersal of seeds. Life cycle of Viscum album is shown in appendix 2. Viscum album is useful in many ways. It is extensively used in the treatment of various illnesses like hypertension, diabetes, anthrosis and cancer (Amer et al., 2012). It provides food for various birds and often animals (Watson, 2001).

1.3. Overview of Viscum album impact on plant species Why Viscum album is a threat to host species? Viscum album has been a threat to its host species and infestation is growing widely in various forests exacerbating loss of tree species. Similarly, Viscum album severely affects the forest, ornamental tree species and orchards in a large extent (Varga et al., 2012). It has been one of the destructive agents for the various plant species. By extracting water and carbohydrate from their hosts Viscum album lead to water and nutrient deficiency (Dobbertin et al., 2005a; Rigling et al., 2010; Varga et al., 2012; Zuber, 2004b). During the drought period, Viscum album effect can be severe to host species as it increases the risk of drought induced mortality of its host species (Rigling et al., 2010). “Viscum album shows much higher transpiration rates and lower water potential than the host trees” (Rigling et al., 2010). Eventually it disrupts the stomatal system including gas exchange effect of the host plant (Zweifel et al., 2012) and thus reducing the photosynthesis phenomenon of the host species (Glatzel & Geils, 2009) Viscum album has been widely distributed in the Alps and dry inner alpine valleys to Bavaria north of the Alps of Europe and has contributed to the decline of pine trees in the forest of Terul (Eastern Spain), Swiss national forest, Germany, Austria, Italy, Greece, Sweden, Great Britain (Dobbertin & Rigling, 2006; Oberhuber, 2001; Peršoh et al., 2010; Sanguesa-Barreda et al., 2012; Tsopelas et al., 2004; Vertui & Tagliaferro, 1998; Zuber, 2004b). Similarly, in Mexican cold mountain forest (evergreen coniferous forest), Viscum album is the second most destructive pest after the bark beetle (Clark-Tapia et al., 2011) . In evergreen coniferous forests of Croatia Viscum album has affected Abies alba (silver fir) in a large extent. A research by Idzojtic et al. (2008) has shown that a high correlation exists between the death rate of Abies alba and presence of Viscum album . Similarly, in Greece , tree mortality was highly correlated with Viscum album infection and 68% of the over storey trees were infected in fir (evergreen coniferous )forest of Mount Parnis (Tsopelas et al., 2004). Various researches have shown that Viscum album infestation is highly associated with higher crown transparency and can be considered as a bio-indicator for tree mortality. According to Varga et al. (2012) huge range of wood species have been infected by Viscum album and it has adversely affected their crown coverage, height and growth rate of the host. In South-western French Alps, the tree mortality of planted Pinus nigra is also in an alarming rate due to Viscum album (Vallauri et al., 2002).

Photo 3: Heavy presence of Viscum album in popular trees (courtesy: Dr. Hein van Gils)


3

1.4. Viscum album impact on Pinus sylvestris Pinus sylvestris has been widely affected by Viscum album and is one of the major causes for its mortality. Viscum album incidence on Pinus species is mainly on Dicrano-Pinion, Erico-Pinion and Ononido-Pinion communities (Zuber, 2004b). Crown degradation, ramification and radial increments are some of the impacts caused by the pine Viscum album to the host Pinus sylvestris (Rigling et al., 2010). The mortality rate of Pinus sylvestris infested by Viscum album is twice higher than that of non-infested trees (Dobbertin & Rigling, 2006). In several studies it was found that the Viscum album infestation is also highly correlated with crown transparency (Dobbertin & Rigling, 2006) and can be considered as one of the bio-indicator of plant species damage and mortality in coniferous forests. Swellings, bending of tree structure, defoliation are some of the impact which can be visually assessed in the Pinus sylvestris with Viscum album. Furthermore, Viscum album considered as an indicator species for changes in temperature and since it extracts the water, it leads to drought stress for the host (Dobbertin et al., 2005b). High mortality rate of Pinus sylvestris due to effect of Viscum album has also been observed in the pine forest of Rhone valley of Switzerland (Dobbertin & Rigling, 2006). Distribution pattern of Pinus sylvestris and Viscum album in Europe including study area is shown in figure1.

[Source: A-(EUFORGEN, 2009) and B-(Zuber, 2004b)]

1.5. Application of remote sensing for tree species classification and pest detection in forestry Remote sensing tools and technologies have been widely used in forestry. Remote sensing is an effective tool for detection, monitoring and mapping of forest health, pest and biological invasive species in forest (Ismail et al., 2007; Joshi et al., 2006). Advancement of high spatial resolution aerial digital imagery (Ortho photo) and satellite imagery has gained importance in accurately assessing and monitoring the health status of the forest. Twenty first century digital aerial imagery has the potential to capture the image at 5 cm spatial resolution and 1 m pixel size (Wulder et al., 2012) and the very high resolution satellite sensors have the ability to take imagery of spatial resolution 50 cm for panchromatic and 2 m for multispectral. Very detailed tree stand level data can be achieved through high resolution digital aerial and satellite imagery

Photo 4: Viscum album in branch of Pinus sylvestris in study area Barcelonnette

Figure 1: Distribution of (A) Pinus sylvestris and (B) Viscum album in Europe


4

(Wulder, 1998; Wulder et al., 2006). According to Hussin et al. (2006) besides high spatial resolution, aerial images (TETRACAM) can show similar spectral characteristics as satellite (IKONOS) imagery can provide. However, both sensors (airborne and satellite) have their unique potential in the classification and monitoring of the forest. Classification and identification of species is in high priority for the conservation and management of forests. Each plant has a unique spectral reflectance and with the development of remote sensing technology and high resolution imageries, it is possible to classify the species with good accuracy (Buddenbaum et al., 2005; Haara & Haarala, 2002; Leckie et al., 2003). Naydenova and Jelev (2009) got the overall classification accuracy of 92% for classifying seven land cover classes using very high spatial resolution aerial photos and satellite images. Dottavio and Williams (1983) were able to delineate heavy defoliation caused by gypsy moth at 88 % accuracy level using Landsat MSS. However, lower accuracies of 65% and 57% were documented for moderate defoliated and non-defoliated hardwood forests in Pennsylvania respectively. Canopy reflectance plays a vital role in the classification of the species. The tree crown is essential when classifying trees from remotely sensed data as its structure and the stand canopy affect the light interceptions which ultimately affect the growth of the tree (Oker-Blom et al., 1986). Moreover, Coops et al. (2006) used high spatial resolution remotely sensed data for the analysis and detection of tree crowns in evergreen coniferous forests subjected to stress and infestation. Furthermore, high spatial resolution multispectral data are capable to detect and map the pest attack with high accuracy (Wulder et al., 2006). Vegetation indices like NDVI (Normalized Difference Vegetation Index) and RGI (Red Green Vegetation Index) could separate the pest attack crowns (red attack) and non-attack crowns (Coops et al., 2006).

1.5.1. Use of spectral reflectance and vegetation indices in monitoring plant health Spectral reflectance and vegetation indices are the key indicators used for assessing and monitoring the health of the plant from remote sensing data. Spectral reflectance of each plant species has its own unique characteristics which vary with wavelength and can be observed in the spectral reflectance plot (Carter & Knapp, 2001). Same species can respond differently according to their health condition (Carter & Knapp, 2001; Xie et al., 2008). “A spectral Vegetation index (VI) is usually a single number derived from the reflectance of two or more wavebands” (Ji & Peters, 2007). Vegetation indices are dimensionless radiometric measures which are computed from reflectance values and widely used to assess the health status of the plant species (Jackson & Huete, 1991). Normalized difference vegetation index (NDVI) is one of the most intensively used spectral vegetation index to analyse the plant stress, biomass, plant health and crop production. NDVI is first used by Rouse et al. (1974), it is calculated on the basis of spectral reflectance in NIR and Red band of the spectrum. Vegetation indices have played a key role in remote sensing based analysis for detecting, mapping and monitoring health status of various trees, forests and invasive plants. The main purpose of the vegetation indices is to enhance the vegetation reflectance/signal and lower the reflectance of other effects like soil and solar irradiance (Jackson & Huete, 1991). Several vegetation indices like NDVI, enhanced vegetation index (EVI) are widely used to assess the health status of plants. The study done by Falkenström and Ekstrand (2002) in an evergreen coniferous forest (Norway spruce and Scots pine) verifies that the analysis of spectral reflectance from high resolution satellite data were able to detect pine defoliation in the near infrared (NIR) band. Furthermore, “high NDVI values would be associated with increased photosynthetic activity and would serve as a useful proxy for identifying stressed trees” (Wulder et al., 2006). Though NDVI shows a greenness (chlorophyll) contents, it gets saturated at high biomass however


5

EVI may address these issues. Wang et al. (2005) emphasized that “EVI is an optimized vegetation index with improved sensitivity in high biomass regions and an improved vegetation monitoring characteristics via a decoupling of the canopy background signal and a reduction in atmospheric influences”. Since the use of red edge in spectral bands, the remote sensing technology has been able to monitor the health status of forest. According to Eitel et al. (2011), red edge plays a crucial role in detecting the stress in plants. The research done on the deciduous broadleaf forest shows that NDVI provides better result in low density canopy whereas EVI provides more accurate result in high density canopy (Ji & Peters, 2007).

1.6. Problem statement and justification Increasing mortality of tree species due to pest infestation is one of the major issues in the conservation and sustainability of the forest. Study of these infected tree species has always intrigued researchers. In the study area (Bois noir catchment, Barcelonnette) Pinus sylvestris as one of the important tree species, is affected by Viscum album in an alarming rate. Viscum album impact has been seen only in Pinus sylvestris in the study area. Conventional approach of fieldwork takes a considerable amount of time and cost. Use of current remote sensing technologies has provided improvements in conventional approaches for field data collection. From coarse resolution satellite data such as Advanced Very High resolution Radiometer (AVHRR) and SPOT VEGETATION to medium resolution imagery (Landsat, Aster or SPOT), are commonly used applications to detect and monitor insect-induced forest disturbances (Blanco et al., 2009; Breshears et al., 2005; Kharuk et al., 2004). Measuring spectral reflectance is one of the methods to distinguish the tree species and it helps to identify the spectral signatures of species. The study area contains coniferous plantation forest with few dominant species and may provide higher classification accuracy. According to Holmgren and Persson (2004), it is possible to detect and distinguish coniferous trees from deciduous trees using near infrared imagery. In addition to this, foliage colour, crown apex, foliage texture, normalized difference vegetation index (NDVI), enhanced vegetation index (EVI) and leaf area index (LAI) are the key variables to identify and detect the health status of tree species. Changes in these indexes will help to assess the health status of the tree species (Coops et al., 2004; Ismail et al., 2008; Leckie et al., 2004). The WorldView-2 (WV-2) satellite is the first high resolution commercial satellite with 8 spectral bands (Digital Globe, 2009). It has the spectral diversity of 4 standard colours (Blue, Green, Red and NIR1) and 4 new colours (Coastal, Yellow, Red Edge and NIR2). The multispectral resolution of this imagery is 1.85 m and it can provide 46 cm panchromatic resolution (Digital Globe, 2009). For the study of vegetation, Yellow, Red edge and Near infrared band plays a crucial role and as a Digital Globe (2009) red edge and NIR able to detect plant stress prominently (Adams et al., 1999). Previous research shows that vegetation indices were good parameters to assess the health status of trees and forest (Leckie et al., 2004; Stone & Coops, 2004). Furthermore, “Red Edge based remote sensing analyses were shown to be effective at identifying trees that were impacted by disease, and were able to provide quantitative information on the health of the trees” (Digital Globe, 2009) . Falkenström and Ekstrand (2002) study in evergreen coniferous forest in Picea abies (Norway Spruce) and Pinus sylvestris verified that, the spectral reflectance from high resolution satellite data can provide pine defoliation in NIR band and also established strong relationship between the ratios of NIR & Red, and NDVI.


6

The research done by Astola et al. (2006) in the boreal forest (evergreen coniferous) of Pinus sylvestris and Picea abies in Eastern Finland shows that, around 88% of trees were correctly classified on the basis of spectral reflectance by using high resolution multispectral imagery (IKONOS). Bhattarai et al. (2011) also found that among several vegetation indices, spectral response of NDVI obtained from WV-2 image is able to classify the Pinus sylvestris infested by Sirex woodwasp. In evergreen broadleaf forest, the classification done on the basis of spectral signature separability and object based image analysis using very high resolution satellite imagery obtained a good accuracy and were able to improve the change detection in species (Chen et al., 2012). Maximum likelihood classifier (MLC) is one of the most commonly and widely used for classification. In MLC, “the decision rule is defined by the multidimensional normal distribution around a class mean” (Liu et al., 2000). The research by Hicke and Logan (2009) got the overall accuracy of 86% and kappa statistics of 0.82 for mapping white-bark pine mortality caused by a mountain pine beetle by using MLC on high spatial resolution satellite imagery. In crop identification, Yang et al. (2011) was able to obtain high accuracy of 91% by applying the MLC method.

Impact of Viscum album in Pinus sylvestris trees in study area (Bois noir, Barcelonnette, France) Photo 5:


7

Pest infestation has been researched using remote sensing technologies; however use of WV-2 satellite imagery for detecting Viscum album infestation in Pinus sylvestris is yet to be studied. This research will help to analyse the possibility and performance of WV-2 for detecting incidence of Viscum album in Pinus sylvestris trees. The generated occurrence of Viscum album in Pinus sylvestris distribution forest map may have added value for the conservation and management of forest in study area.

1.7. Research objectives: The overall goal of the research is to detect and map the presence of Viscum album in Pinus sylvestris forest using very high resolution satellite imagery (WV-2) and multispectral Ortho-photo of airborne digital camera.

1.7.1. Specific objectives: 1. To accurately classify the Pinus sylvestris forest using WV-2 image. 2. To study the spectral distinctness of Pinus sylvestris in the presence and absence of Viscum album. 3. To determine the relationship of presence of Viscum album in Pinus sylvestris with elevation and

DBH. 4. To determine whether very high resolution multispectral imagery can detect the presence of

Viscum album in Pinus sylvestris.

1.8. Research questions 1. Is Pinus sylvestris with and without Viscum album is spectrally distinct? 2. Can vegetation Indices (NDVI-with different band combination) derived from multispectral

imagery (WV-2) differentiate the presence and absence of Viscum album in Pinus sylvestris? 3. Is there any significant relationship in incidence of Viscum album in Pinus sylvestris with elevation

and DBH? 4. How accurately can the incidence of Viscum album in Pinus sylvestris forest be detected by the use of

maximum likelihood classification (pixel based classification) on WV-2 satellite and Ortho imagery?

1.9. Hypothesis: 1. H1: There is a significant difference in the spectral reflectance of Pinus sylvestris in the presence and

absence of Viscum album. H2: There is no significant difference in the vegetation indices of Pinus sylvestris in the presence and absence of Viscum album.

2. H1: There is no significant difference in the incidence of Viscum album with elevation. H2: High incidence of Viscum album in low elevation.

3. H1: can be detected Viscum album in Pinus sylvestris forest with significant accuracy by applying pixel based MLC on high resolution imagery (WV-2).

1.10. Research innovation The innovation of this research is aimed at detection of Viscum album in Pinus sylvestris tree by image classification on WV-2 satellite imagery which has not yet been done. This research will further add value in application of remote sensing and the use of high resolution imagery (WV-2) in detecting the Viscum album infestation in forest and will eventually help in forest conservation and management.


8

2. DESCRIPTION OF STUDY AREA 2.1. Background

Barcelonnette is a small municipality situated in the South Western French Alps. Barcelonnette is also known as the heart of the Ubaye Valley and a home for around 3500 people (www.barcelonnette.com). Sunny days, clear cold nights are the climatic features of the area. Barcelonnette is a major tourist attraction for the skiing, biking, hiking, paragliding and rafting. 2.2. Study Area

The study area is a part of Bois noir catchment (3 km2) with the elevation ranges from 1100 to 3000 m. However, the core study area is 0.6 km2 and located in south and east part of the catchment between 1402 to 1700 m. Location map of the study area is given in appendix 3.

2.2.1. Climate The climate of the study area is dry mountainous. Inter annual rainfall strongly varies from 400 mm to 1400 mm and the mean annual temperature is around 7.5°c (Maquaire et al., 2003).

2.2.2. Geology: The area has an irregular rugged topography with slope gradients ranging from 10° and 35° and scree slopes (Saez et al., 2012; Thiery et al., 2007). The study area is covered by 15 m thick Morainic colluvium, underlain by autochtonous Callovo –Oxfordian black marls and both materials are highly sensitive to landslides (Maquaire et al., 2003). The area is highly landslides prone zone and mostly planted by pine trees in 19th century. Recently subject to multiple reactivations of landslides (Razak et al., 2011b) and a rotational slide is quite prominent (Saez et al., 2012).

Photo 6: Overview of the study area -Southern hills (Bois noir, Barcelonnette, France)


9

2.2.3. Vegetation The vegetation of the study area is a century old plantation of coniferous trees with pockets of broadleaf forest. In 18th and 19th century, the Bois noir area suffered from heavy deforestation (Kappes et al., 2011). To control deforestation and debris flow, reforestation and construction of check dams was initiated (Kappes et al., 2011). At present, 92% of area is covered by forest (Thiery et al., 2007). Pinus sylvestris, Pinus uncinata, Picea abies and Larix decidua Miller are the major tree species. These plantations are also in the abandoned agricultural fields and the evidence of remnants of terraced crop farming has been observed. According to Saez et al. (2012) average mean age of trees in the Bois noir catchment is of 100 years. High and frequent landslide activities in the Bois noir catchment has disrupted the tree stand structures often known as “drunken trees” (Razak et al., 2011a). Species identification and its description were recorded from Dr. Hein A.M.J. van Gils (2011, 2012).

(Source: Photo 6 & 7 -Chaudon Norante,)

2.2.3.1. Pinus sylvestris L. (Le Pin sylvestre): Pinus sylvestris is an indigenous species in the dry inner alpine valleys and the dry Alps, distribution shown in figure 1. Pinus sylvestris have their own cuticle illustration of needles and stomata which clearly distinguished (Fauvart et al., 2012) from other coniferous species like mountain pine (Pinus uncinata). The upper half of the stem is flaky and orange in color whereas the lower scaly dark grey brown color and the cones are symmetrical with an umbo cantered. Some of the patches of Pinus sylvestris tree structures are highly disrupted. This may be due to active and frequent landslides in the study area.

Photo 7: Deforestation (1905 AD)

Photo 8: Coniferous plantation (1997 AD)


10

Figure 2: Physical structure of cones of Pinus sylvestris, Pinus uncinata and Pinus mugo

[Source: (Fauvart et al., 2012)]

2.2.3.2. Mountain pine [Pinus uncinata Mill. ex Mirb. (Le Pin a crochet)]: Pinus uncinata is also called Swiss Mountain pine which naturally found at the tree line, in Pyrenees and the Western Alps. It is sometime conisdered a sub species of Pinus mugo and aslo found in the central to Eastern Alps, the Balkan and Apennins. Its cones are assymetrical and have a hook-shaped umbo with much thicker scales on one side of the cone which makes it distinguished from other conferous trees. Pinus uncinata can grow from 12 to 20 m tall. Pinus uncinata also has been used in France for land rehabilitation and controlling of landslides. In research area, Pinus uncinata trees age varies from 60 to 80 years and are younger compared to other tree species (Saez et al., 2012). High numbers of disrupted Pinus uncinata (drunken trees) are found in the upper part of the study area.

2.2.3.3. Larix decidua Miller.: Larix decidua commonly known as larch is a deciduous conferous tree. It is indigenous in the Alps, central Europe and Carpathians.The size of Larix decidua varies but can grow up to 45 m tall.

2.2.3.4. Picea abies L. (Karsten): Picea abies, an evergreen coniferous trees belongs to spruce species and is native to Europe. It is a fast growing species and can grow upto 55m tall. This species is also widely planted outside its native habitat.

Photo 8 : “ Drunken” Pinus sylvestris trees Photo 9 : “ Drunken” Pinus uncinata trees


11

3. MATERIALS AND METHODS 3.1. Satellite data

Very high resolution satellite imagery (WV-2) of having 8 bands spectral resolution of 2m and panchromatic resolution of 0.5m was used for this study. The imagery was acquired in September 2010 and the metadata of this imagery is given in appendix 4.

3.2. Digital aerial Ortho image and LiDAR data The aerial Ortho- photo and airborne LiDAR data acquired in July 2009 under snow free conditions were used. The dataset was initially collected for the study of landslides activities and the aerial ortho image having 15 cm resolution was also captured at the same time. Aerial Ortho image consists of 3 bands (red, green, blue). The LiDAR derived canopy height model (CHM) and digital terrain model (DTM) were used as explained in section 3.5.2 .

3.3. Field material Following listed equipment were used to collect data for fieldwork.

List of equipment Table 1.

SN

Equipment Purpose

1 iPAQ and Garmin GPS Navigation

2 Leica differential GPS system 1200 Recording spatial co-ordinates 3 Sunnto compass Orientation 4 Diameter tape (5m) Measurement of DBH 5 Measuring tape (50 m) Measuring radius of circular plot and length of

transect plot 6 Spherical densitometer Measuring canopy density 7 Fieldwork datasheet Field data record

3.4. Processing Software Following listed software were used for this research.

List of software and its purpose of usage Table 2.

SN

Software Purpose of Usage

1 ArcGIS2010 GIS Analysis

2 Erdas Imagine 2011 and ENVI 4.7.2 Image Processing

3 SPSS and R Statistical Analysis

4 Word, Excel, Powerpoint, End note

and Visio

Thesis Writing and editing


12

3.5. Methods The research methodology is shown in figure 4.

3.5.1. Sampling Design The field sampling design was based on two methods. Circular plot and transect samplings were adopted for data collection in the field. To avoid bias, a random set of 40 location points were generated in the study area by using a GIS. Only 31 circular plots were surveyed. So as to increase the Viscum album incidence data 27 transects were sampled as further explained in section 3.5.2 on field techniques. Due to time limitations and topographic constraints 7 sample points were discarded and 2 plots without Pinus sylvestris trees were also removed from the sampling set.

Figure 3: Flowchart of research


13

Transect and circular sampling methods were applied due to following reasons:

Circular plot were used to assess the forest structure parameters like tree density, abundance, frequency and species diversity analysis (Boakye et al., 2012).

Circular plot has a minimum perimeter with no fixed orientation and easy to measure in the forest (Husch et al., 2003)

Line transect method is widely used in forest data collection including biological invasion of forest (Günter et al., 2007; van Gils et al., 2004; van Gils et al., 2006) .

Most of the statistical analysis was performed on the basis of circular plot data whereas transect data were added for training samples and used in accuracy assessment.

3.5.2. Field techniques: Field data collection was carried out from 11th September to 27th of September 2012 during leaf on conditions. The iPAQ and Garmin GPS were used to navigate to the selected plot centre. However, due to canopy obstructions it was impossible to locate plot centres precisely. To minimize the error, the available LiDAR derived canopy height model (CHM) (Kumar, 2012) was used to determine the centre of the plot. Using the CHM, at least two landmark features were identified in the surroundings of centre plot as seen on the CHM to confirm plot centres. The centre point and individual locations of trees were subsequently determined by measuring the bearing and distance using the sunnto compass and measuring tape respectively. Differential Global Positioning System (DGPS) was also used to record the geographic coordinate of centre point of the plot. From the centre, the circular plot having 500 m2 areas was established to collect the data after slope correction (Husch et al., 2003) . Within a plot, all trees species were counted and Pinus sylvestris trees with diameter at breast height (DBH) greater than 3 cm were measured using a DBH tape. Each Pinus sylvestris tree in the plots were visually assessed on the ground and recorded the incidence of Viscum album. Crown coverage of the plot was measured using a spherical densiometer from five different locations within the plot and canopy cover was averaged. Photographs of each plot were taken at the same time in the field to recognize the trees in the WV-2 image and Ortho photo. After sampling of a circular plot, four 50 m perpendicular line transects were laid from the same centre point. The presence (P) and absence (A) of Viscum album in Pinus sylvestris trees was recorded at 5 m intervals. The XY co-ordinate, type of forest and density of Viscum album of each Pinus sylvestris tree species was recorded and observed on the ground. Density of Viscum album in Pinus sylvestris was given as follows: 0=No Viscum album 1=Low Viscum album 2=Medium Viscum album 3=High Viscum album


14

Figure 4: Field sample points locations in Ortho photo


15

3.6. Image pre-processing The WorldView-2 (WV-2) image was delivered without any cloud cover, and both atmospherically and radiometrically corrected. The image was pre-processed in three stages; (i) geometric correction, (ii) pan sharpening and (ii) image enhancement.

3.6.1. Geometric correction In the first stage, the aerial Ortho photo captured during the collection of LiDAR data was geo-referenced. However there was a challenge to match the high resolution WV-2 optical imagery with LiDAR derived CHM and aerial Ortho photo. This is due to differences in geometric properties. WV-2 rational polynomial coefficients model available in ERDAS© software was used to transform the image coordinates to earth surface coordinates. Rational Polynomial Coefficients (RPC) is an empirical mathematical model which is relating the image space to latitude, longitude and surface elevation. In the second stage, six ground control points were used to improve the geometric precision between the optical imagery and provided LiDAR derived surfaces. Identical areas (house and corner of the road) in the optical and aerial images were tied in ERDAS LPS tools. Both panchromatic and multispectral WV-2 bands were shifted 12m with an average root mean square error of 1 pixel.

3.6.2. Image fusion and enhancement WV-2 multispectral image of 2m resolution and 0.5m resolution of WV-2 panchromatic image were fused through HCS pan sharpening algorithm by using ERDAS© software as described by Padwick et al. (2010). This algorithm is specifically developed for WV-2 images and has the capability of maintaining the spatial and spectral recovery of all bands. For image enhancement, five different window filter size were applied. In HCS resolution merge it has the smoothing filters of five different sizes (3x3, 5x5, 7x7, 9x9 and 11x11) were applied on the image to cross check the image quality before finalizing the pan sharpened image. There was no significant difference in spectral and spatial quality of the five images obtained using different convolution filtering sizes. To choose the best pan sharpened image, the visual interpretation was applied and the image acquired through using the dimension of 7x7 convolution filter was chosen.

3.7. Fieldwork data analysis Normality of the field data was checked. The descriptive statistical analysis (Husch et al., 2003; Zobel et al., 1985) of the field data was done. The tree and forest variables were viewed in histograms and whisker boxplots were prepared using R software to depict the presence and absence of Viscum album in Pinus sylvestris as with the elevation and DBH.

3.7.1. Vegetation analysis The following parameters were calculated for the forest analysis.

Equation (1)

Equation (2)

Equation (3)


16

Equation (4)

Equation (5)

(cm2) Equation (6)

Where, DBH = diameter of tree at breast height

3.7.2. Tree diversity analysis:

Shannon diversity index, species richness (no. of species) and species evenness (Magurran, 1988; Margalef, 1958; Shannon & Weaver, 1962) were calculated for each plot. The following equations were applied for the calculation of these indexes Shannon Diversity Index (H’):

Equation (7) Where pi = the proportion of individuals belonging to ith tree species Species richness:

Equation (8)

Where, S = total number of species N = total number of individuals in the sample Species Evenness:

Equation (9)

H = Shannon–Wiener diversity index S = total number of species in the sample

3.7.3. Statistical analysis The t-test, correlation analysis, ANOVA test and regression analysis were used. Field data of presence and absence of Viscum album was tested by t-test to determine the relationship with elevation and DBH. Furthermore, two models were developed using R software to estimate the presence of Viscum album in Pinus sylvestris by applying multiple regression and logistic regression method. Elevation and DBH were used as explanatory variables. The binomial logistic regression model was selected in this research. The dependent variable (presence and absence of Viscum album) are dichotomous and the independent variables (DBH and elevation) are used as predictors. A value 1 corresponds to the presence of Viscum album whereas a value 0 characterizes the absence of Viscum album in Pinus sylvestris. The model was performed by using R-statistical package. The equations used for the model were as follows: Multiple Regression:


17

Equation (10) For Logistic regression:

Equation (11) Where, b0 is slope intercepts and b1 and b2 are estimated parameters for DBH and elevation in both models.

3.7.4. Image classification At first the WV-2 image was classified into the major dominant tree species by applying supervised MLC through ERDAS© software. From this image, only Pinus sylvestris forests were extracted by masking in ArcGIS 10 software in both the processed WV-2 image and Ortho image. Following are the steps applied for this classification:

3.7.4.1. Supervised pixel based maximum likelihood classification (MLC): For decades, MLC has been a most widespread classification method in remote sensing. MLC is a statistical classification method which uses both the mean and centre of each class that relies on Gaussian distribution and a pixel with a maximum likelihood was considered as a same class (Kavzoglu & Reis, 2008). While assigning classes, both variance and covariance of the class signatures are considered. The field training data was used to collect a signature to classify the pixel as a presence and absence of Viscum album in pine tree/forest by using supervised MLC. The shapes, sizes, distance and locations are considered while assigning the pixels to each class and “ if the assumption of a normal distribution for each class is correct then the classification has a minimum overall probability of error and the MLC is the optimal choice” (Kavzoglu & Reis, 2008). Since the Viscum album is limited to Pinus sylvestris only three classes namely; (1) Pinus sylvestris, (2) broadleaf and (3) others were classified by applying supervised MLC algorithm through ERDAS© software. Five reference point (co-ordinates) for the class ‘others’ were manually picked from Ortho image to increase a sufficient number of reference point to test the classification accuracy. In the second stage, the Pinus sylvestris data obtained from the classification was exported using ArcGIS to create only Pinus sylvestris data. And both the images (WV-2 and Ortho image) were masked by acquired Pinus sylvestris data through ArcGIS software and the distribution map of Pinus sylvestris was generated.

3.7.5. Extraction of canopy spectra of Pinus sylvestris in presence and absence of Viscum album from WV-2 image:

Training samples of presence and absence of Viscum album in Pinus sylvestris were obtained using field observations as a reference. Pixel value of sample Pinus sylvestris trees was extracted from the spectral profile and exported to excel sheet. ERDAS© software was applied for the extraction of canopy spectra. The mean pixel values (DN value) of Pinus sylvestris trees with and without Viscum album were plotted in Y axis and the eight spectral bands of WV-2 were plotted in the X-axis.

3.7.6. Computing and analysis of spectral vegetation indices Three equations using different band combinations for NDVI (Rouse et al., 1974) were computed (table 3) in the model maker of ERDAS© software. Three different NDVI maps were generated. Field points of presence and absence of Viscum album were overlaid on the three NDVI images. To extract pixel values from the NDVI images for spectral signature development an inquiry box of minimum of 6 pixels was


18

used. The pixels values at the determined geographic location were exported to ASCII file. The mean value of those exported presence and absence of Viscum album in Pinus sylvestris was plotted in Excel and the analysis was carried out.

NDVI Vegetation Indices Table 3.

S.N. Abbreviation 1 NDVI 76 2 NDVI 85 3 NDVI 86

Where, 5= Band 5 (Red) 6= Band 6 (Red edge) 7= Band 7 (NIR1) 8= Band 8 (NIR2) (*Note: Cross reference for wavelength range, see figure 22) The two sample t-test was used to analyse the performance of each vegetation index for distinguishing the presence and absence of Viscum album in Pinus sylvestris. The performance of different vegetation indices to discriminate the presence and absence of Viscum album was compared using the Kruskal-Wallis test (Kruskal and Wallis, 1952). If the Kruskal-Wallis test indicated a significant difference, the Mann-Whitney U-test with Bonferroni correction was applied to account for the multiple comparisons among the different vegetation indices.

3.7.7. Supervised pixel based maximum likelihood classification (MLC) of WV-2 and Ortho image: In supervised classification it is necessary to know the spectral characteristics of each class to make a decision rule to classify the whole population. For the classification of presence and absence of Viscum album in Pinus sylvestris, once again supervised MLC was used in ERDAS© software and the distribution map of presence of Viscum album in Pinus sylvestris were generated.

3.8. Accuracy assessment: The accuracy of all three classified map was assessed on the basis of a confusion matrix and Cohen’s kappa statistic (Congalton, 1991). Confusion matrix is a specific table having the quantitative or statistical information regarding actual and predicted classifications obtained through the classification algorithm/system. Confusion matrix records the major four parameters (i) true positive, (ii) false positive, (iii) false negative and (iv) true negative (Allouche et al., 2006) to evaluate the classification accuracy. Kappa statistics is the accuracy measurement methods which is based on the difference between observed agreement and expected agreement (Viera & Garrett, 2005). The kappa coefficient summarizes an agreement table (Sim & Wright, 2005; Warrens, 2011) and kappa coefficient value ranges from negative 1 to positive 1. Higher value signifies the better performance whereas lower value represents the poor performance. Negative 1 indicates a complete disagreement between observed and expected agreement whereas positive 1 signifies a prefect agreement. Landis and Koch (1977) assess the Cohen’s value as excellent (kappa>0.75), fair to good (0.75>kappa>0.4) and poor (kappa<0.4). Kappa statistics can be calculated as:

Equation (12)

Where,


19

Po= Observed agreement Pe= Expected agreement

The true skill statistic (TSS) of each classified image was also calculated to assess the accuracy of image classification. Even though the kappa is widely used in accuracy or model assessment, it has some shortcomings which are compensated by TSS measure (Allouche et al., 2006). The value of TSS also ranges from negative 1 to positive 1. The TSS is calculated as follows:

Equation (13)

Equation (14)

Equation (15)


20

4. RESULTS

4.1. Tree species in the forest The study area is dominated by Pinus sylvestris. However seven other woody species were recorded. The proportion of Pinus sylvestris is more than 75% in 60% of the analysed plots. In 15% of plots the ratio of Pinus sylvestris species ranges from 30 to 60%. Broadleaf covers around 11% of the forest area and followed by mountain pine (Pinus uncinata) by 8% and Alnus viridis covers the least (0.5%).

Among the eight tree species, Pinus sylvestris has the highest density of around 617 trees/ ha whereas Alnus viridis and Picea abies less than 5 trees/ ha. Descriptive analysis is given in appendix 8.

4.2. Tree diversity: About 16% of the plot contains only Pinus sylvestris trees and 74 % of the sampled plot having canopy cover of 50 to 75%. Similarly 13% of the plots have canopy coverage of less than 50% and more than 75% respectively. Within the plots, the Shannon diversity index ranges from 0 to 1.3. The average Shannon diversity index is 0.9 and species richness and evenness is 0.97 and 0.43 respectively.

0

100

200

300

400

500

600

700

Alnusviridis

Broadleaf Fraxinusexcelsior

Larixdecidua

Piceaabies

Pinussylvestris

Pinusuncinata

Populustremula

Dens

ity (N

o./h

a)

Figure 6: Density of tree species in sampled plot

0 20 40 60 80 100

Alnus viridisBroadleaf

Fraxinus excelsiorLarix decidua

Picea abiesPinus sylvestrisPinus uncinata

Populus tremula

Frequency %

Figure 5: Frequency of tree species in sampled plot


21

4.3. Descriptive statisics of Pinus sylvestris forest Out of 1276 sampled trees 956 are Pinus sylvestris. Histograms of elevation, DBH and height are shown in appendix 9. DBH of Pinus sylvestris ranges from 3 to 59cm with an average DBH of 18cm and the basal area coverage is 20.83 m2/ ha. Similarly, maximum average height of the sampled tree was found to be 23 m with average height of 13m. In study area Pinus sylvestris trees was found within the range of 1420 to 1600m. More than 70% of sampled Pinus sylvestris tree DBH class belongs to pole size (DBH=10 to 30cm) and 17% and 12% of tree belongs to sapling (DBh <10cm) and sawlog category (>30cm).General descriptive statistics of Pinus sylvestris tree parameters are given in appendix 9.

4.3.1. Relationship of DBH and height Scatter plot shows that there is a weak relationship between the height and DBH of Pinus sylvestris trees. There is a minimum or no increase in DBH as height increases. The regression equation and obtained R2 value is shown in the figure 10.

y = 0.1717x + 8.7317R² = 0.2567

0

5

10

15

20

25

0 20 40 60 80

Heig

ht (m

)

DBH (cm)

Figure 7: Pinus sylvestris tree distribution parameters (A) DBH, (B) Elevation, (C) Height

Figure 9: Scatter plot showing relationship of DBH and height of Pinus sylvestris

Figure 8: Pinus sylvestris tree parameters (A) DBH, (B) Elevation, (C) Height


22

4.4. Descriptive anlysis of presence of Viscum album in Pinus sylvestris Out of 31 sampled plots, 17 plots have presence of Pinus sylvestris. Descriptive analysis of density, relative density, frequency, relative frequency, abundance, basal area and relative basal area was presented in appendix 12. Within the Viscum album affected Pinus sylvestris trees, less than one third of Pinus sylvestris have high density of Viscum album which is shown in figure 11.

The overall density of Pinus sylvestris tree is 618 trees/ha whereas the density of Pinus sylvestris having Viscum album is 40 trees/ha. Most of the affected Pinus sylvestris trees have medium level of Viscum album which accounts around 16 trees/ha and almost equal number of trees has presence of high and low (13 and 12 trees/ha respectively) density of Viscum album. Around 3.5 ha of basal area are covered by Pinus sylvestris trees having Viscum album.

Figure 13: Basal area and of Pinus sylvestris having different level of Viscum album

The relative frequency of Pinus sylvestris trees with low density Viscum album is almost half of the total affected trees whereas nearly equal distribution of high and medium density level of Viscum album in Pinus sylvestris is found in the study area. However, the relative abundance of low density Viscum album in Pinus sylvestris trees is less than other two classes.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

High Medium Low

m²/

ha

Viscum album level in Pinus sylvestris

Basal Area

55%

45%

Presence

Absence

Figure 11: Incidence of Viscum album in Pinus sylvestris trees

27%

31%

42%

High

Medium

Low

Figure 10: Relative frequency of Pinus sylvestris having different density level of Viscum album

02468

1012141618

High Medium Low

No.

/ha

Viscum album level in Pinus sylvestris

Density

Figure 12: Density of Pinus sylvestris having different level of Viscum album


23

4.5. Relationship of Viscum album in Pinus sylvestris with elevation and DBH Presence and absence of Viscum album in Pinus sylvestris is equally distributed (around 6%) in lower elevation ranges from 1420 to 1500m. High percentage (36%) of presence of Viscum album was found between the elevation ranges of 1500 to 1550m. However, in higher elevation the presence of Viscum album is lower in Pinus sylvestris trees.

0

5

10

15

20

25

30

35

40

45

Relative Frequency Relative Abundance

%

High

Medium

Low

Figure 14: Relative frequency and relative abundance of Pinus sylvestris having different density class of Viscum album

Figure 15: Incidence of Viscum album and elevation (m) (plots)

0

10

20

30

40

50

60

70

80

90

1420-1500 1500-1550 1550-1600

% o

f inc

iden

ce o

f Visc

um a

lbum

Elevation Range (m)

Presence

Absence

Figure 16: Incidence of Viscum album in different elevation ranges (plots)


24

The presence of Viscum album is low in Pinus sylvestris having low DBH and big DBH have high presence of Viscum album. In sampled trees, 10% of Pinus sylvestris trees have presence of Viscum album having DBH more than 30cm and only 0.3% of Pinus sylvestris have occurrence of Viscum album having DBH less than 15cm.

Pinus sylvestris having DBH less than 15cm has only low density level of Viscum album where as DBH greater than 30cm have all three density (high, medium,low) level of Viscum album and contributing 18% of presence of high density Viscum album.

A negative correlation between the Viscum album incidences on Pinus sylvestris and elevation (r=-0.52; p<0.001). However there was a positive significant correlation between the Viscum album incidences on

Pinus sylvestris and DBH (r=0.52; p<0.001). Both T-test and ANOVA test ay 95% confidence level showed

(table 4 and 5) the significant relationship between presence of Viscum album in Pinus sylvestris with

elevation and DBH.

0

5

10

15

20

25

<15 16-20 20-30 >30

% o

f in

cide

nce

of V

iscu

m

albu

m

DBH (cm)

high

Medium

Low

0

5

10

15

20

25

30

35

<15 16-20 20-30 >30

% o

f in

cide

nce

of V

iscu

m a

lbum

DBH (cm)

Presence

Absence

Figure 17: Presence of Viscum album in Pinus sylvestris per DBH class (plots)

Figure 18: Boxplot of Pinus sylvestris DBH of presence/absence of Viscum album (plots)

Figure 19: Percentage of occurrence of different density level of Viscum album on different DBH classes of Pinus sylvestris


25

T-test for determining relationship of presence of Viscum album with elevation and DBH Table 4.

Variables d.f. t-stat t-Critical p-value Pearson correlation (r) Conf. interval Elevation 54 7.25 1.6736 1.60E-09 -0.5135 95% DBH 54 7.3663 1.6736 1.05E-09 0.5273 95%

ANOVA for determining relationship of presence of Viscum album with elevation and DBH Table 5.

Variables d.f. F-Stat F Critical p-value Conf. interval Elevation 317 20.39 3.87 0.00 95% DBH 315 41.59 3.87 0.00 95%

4.6. Multiple regression The model developed through multiple regression method was tested using t-test and F-test. Though, R2 is value is only 0.4453 the model is statistically significant at 95% confidence level. A summary of the test is given in table (6 & 7) and the equation of the model is as follows:

Y=8.4865 + (-0.00567) *(Elevation) + 0.0248*DBH

Summary of multiple regression for prediction of Viscum album in Pinus sylvestris trees Table 6.

Variable Estimate Standard Error t Pr>|t| Significances at alpha=0.5 Intercept 8.4865636 1.2523623 6.776 1.96E-10 Very significant Elevation -0.00567 0.0007945 -7.136 2.69E-11 Very significant DBH 0.0248498 0.003341 7.438 4.92E-12 Very significant

Summary of goodness of fit Table 7.

Test Statistics/Parameters Value R2 0.4453 Adjusted R2 0.4387 F-statistic 67.83 on 2 and 169 DF p-value < 2.2e-16

4.7. Logistic regression Stepwise logistic regression was performed to estimate the best fit model. The model with both explanatory variables Elevation and DBH result in the lower AIC value (130) and the most significant model was chosen. The Nagaelkerke R-square value for this model is 0.58 and the equation is as follows:

Y=1(1+exp (-1*(63.5+0.18*DBH-0.17*Elevation)))

AIC value of models Table 8.

Model No. Explanatory Variables AIC 1 Elevation, DBH 130.05 2 Elevation 168 3 DBH 169.19

Both predictors are highly significant. Coefficients and significances of predictors of the best logistic regression model are shown in table 9.


26

Coefficients and significances of predictors Table 9.

Predictors Estimate Std. Error z value Pr(>|z|) Significances at alpha=0.5

Intercept 63.5 13.67 4.65 3.35e-06 Very significant DBH 0.18 0.033 5.36 8.09e-08 Very significant Elevation -0.045 0.009 -4.96 7.12e-07 Very significant

4.8. Image processing WV-2 image and Ortho photo were considerably matched after the geometric correction. The WV-2 image was shifted 12m from its original location. For instance, a house was matched in both images (Ortho and WV-2) which are shown in figure 20.

4.9. Image classification High overall classification accuracy (96%) was achieved for the mapping of distribution of Pinus sylvestris. Among three classes namely (i) Pinus sylvestris, (ii) broadleaf and (iii) others, the user accuracy is higher for Pinus sylvestris (97%) and least for the class others. Similarly, the overall Kappa value of 0.84 was obtained for tree species classification.

Figure 20: Image location showing 'before' and 'after' geometric correction


27

Species classification accuracy assessment Table 10.

Species

Reference Error Commission %

User Accuracy % Pinus sylvestris Broadleaf Others Total

Clas

sifie

d

Pinus sylvestris 220 4 2 226 2.65 97.35 Broadleaf 0 16 1 17 5.88 94.12 Others 3 1 17 21 19.05 80.95 Total 223 21 20 264

Error Omission 1.34 23.81 15 Overall Accuracy (%) 95.83

Producer Accuracy %

98.65 76.19 85

Figure 21: Distribution map of Pinus sylvestris in Bois noir, Barcelonnette France (WV-2 image)


28

4.10. Spectral analysis of Pinus sylvestris in the presence and absence of Viscum album Spectral distinctness of Pinus sylvestris in the presence and absence of Viscum album is clearly observed in all eight bands of the WV-2 image. All band showed a similar trend of spectral curve for both presence and absence of Viscum album in Pinus sylvestris. Furthermore, Pinus sylvestris in the presence of V. album has a low spectral reflectance in all bands. However, NIR1, NIR2 and red edge have the highest spectral distinctness compared to other bands. In the red edge, there is a sharp rise in the reflectance curve in both presence and absence of Viscum album. However the presence of Viscum album shows a short sharp rise compared to Pinus sylvestris in the absence of Viscum album. Spectral reflectance of Pinus with and without Viscum album is shown below:

4.11. Analysis of vegetation indices (NDVIs) The mean NDVI pixel value of presence and absence of Viscum album was calculated which is given in table 4-8. In all three calculated indices, the NDVI value is lower in presence of Viscum album in Pinus sylvestris. However, NDVI value (presence=0.61, absence =0.70) obtained from band combination of 8 and 5 (NIR2 & red band) is able to detect the presence of Viscum album better compared to other two indices.

Reflectance value of three vegetation indices Table 11.

NDVI

Viscum album Presence Absence

76 0.090 0.097 85 0.619 0.704 86 0.100 0.155

NDVI 85 was significant at (alpha=0.05, p-value<0.00) for discriminating the presence and absence of Viscum album in Pinus sylvestris. However, NDVI 76 was not significant at alpha of 0.05. Furthermore, a Kruskal-Wallis test showed NDVI 85 was significantly different (p<0.0167) from NDVI 76 and NDVI 85. NDVI 85 had highest discrimination and was significantly different from other two vegetation indices.

0

100

200

300

400

500

600

700

800

400-450(Coastal

Blue)

450-510(Blue)

510-580(Green)

585-625(Yellow)

630-690(Red)

705-745(Red Edge)

770-895(NIR 1)

860-1040(NIR 2)

Spec

tral

Ref

lect

ance

(nm

)

Wavelength in nm

AbsencePresence

Figure 22: Spectral reflectance of Pinus sylvestris in the presence and absence of Viscum album


29

4.12. Classification of presence and absence of Viscum album in Pinus sylvestris in the WV-2 image The WV-2 image was successfully classified to detect the presence and absence of Viscum album in Pinus sylvestris by applying pixel based MLC. The overall classification accuracy of 86% with overall kappa value of 0.52 was achieved for detection of presence of Viscum album in Pinus sylvestris forest. The user accuracy of 72% was achieved for presence of Viscum album and 88% for absence of Viscum album. The kappa (K^) statistics for presence and absence of Viscum album are given in table 12 and 13. Furthermore, the true skill statistic value of 0.6 was obtained for the detection of V. album. The error of commission and omission for the presence of Viscum album is higher (28% and 48% respectively) than the absence of Viscum album. Even though the overall accuracy is higher, the false presence rate is also higher 0.48 compared to false absence rate. Similarly, sensitivity and specificity of the classified image was obtained 0.72 and 0.88 with true skill statistics value of 0.6.

Accuracy assessment for the WV-2 image for presence and absence of Viscum album in Pinus Table 12.sylvestris

Status Reference Error

Commission % UserAccuracy % Presence Absence Total

Clas

sifie

d Presence 13 5 18 27.78 72..2 Absence 12 93 105 11.43 88.5

Total 25 98 123

Error Omission 48 5.1 Overall Classification Accuracy (%)

86.2 Producer Accuracy %

52 94

Kappa value for presence and absence of Viscum album for classified WV-2 image Table 13.

Status Kappa (K^) Presence of Viscum album 0.6514 Absence of Viscum album 0.4377

Figure 23: Boxplot of reflectance value of three different vegetation indices for presence and absence of Viscum album in Pinus sylvestris


30

Figure 24: Distribution map (WV-2) of presence of Viscum album in Pinus sylvestris forest


31

4.13. Classification of presence and absence of Viscum album in Pinus sylvestris on airborne multispectral Ortho image

The overall classification accuracy of 77.7% with overall kappa value of 0.3662 was obtained for detection of presence and absence of Viscum album in Pinus sylvestris trees. The kappa (K^) statistics for individual species is given in table 4-11 and 4-12. Similarly, sensitivity and specificity of the classified image was obtained 0.71 and 0.78 with true skill statistics value of 0.5.

Accuracy assessment for Ortho image for presence and absence of Viscum album in Pinus sylvestris Table 14.

Status

Reference Error Commission %

User Accuracy %

Presence Absence Total

Clas

sifie

d Presence 15 6 21 28.6 71.43 Absence 25 93 118 21.18 78.81

Total 40 99 139

Error Omission 62.5 6.06 Overall Accuracy (%) 77.7 Producer Accuracy % 37.5 93.94

Kappa value for presence and absence of Viscum album for classified Ortho image Table 15.

Status Kappa (K^) Presence of Viscum album 0.5988 Absence of Viscum album 0.2638


32

Figure 25: Distribution map (Ortho image) of presence of Viscum album in Pinus sylvestris forest


33

5. DISCUSSION

5.1. Pinus sylvestris plantation forest Low tree diversity combined with high tree density is the characteristics of plantation forest (Hartley, 2002). The Pinus sylvestris tree density (823/ha) found in the study area is six times higher than that of (132/ha) the natural pine forest (Summers et al., 1999). Pole size Pinus sylvestris dominated in the Bois noir whereas the stem size is more diverse in natural pine forest (Kint, 2005; Summers et al., 1999). Furthermore, Watt and Kirschbaum (2011) found that the tree DBH and height of even-aged plantations coniferous have a linear relationship (0.73 R2). However, in our area Pinus sylvestris DBH and height show a weak relationship (0.25 R2). Similarly, the obtained Shannon index (0.9), species richness (0.97) and evenness (0.43) show low tree diversity compared to results obtained by Lilja and Kuuluvainen (2005) for near natural Pinus sylvestris forest (Shannon index of 1.88). Similarly, Pourababaei et al. (2012) found the species richness of 1.71 for coniferous plantation. According to Zhang et al. (2012) higher species richness and evenness can improve the productivity and diversity of the forest. In plantation forest, understory vegetation were reduced by the dense stand and adversely affect the native species (Gómez-Aparicio et al., 2009). However, the diversity and species richness increases in plantation forest in less than a century (Lust et al., 1998). Although the plantation forest is more than 100 years old, the species diversity is low and the tree DBH is not significantly correlated with height. The growth rate of the tree is significantly lower in Bois noir (Saez et al.). Similarly, the reason for low diversity could be the lack of relict montane conifer forest. Furthermore, there is no thinning and harvesting of trees and the reason could be the inaccessibility of the sites for the timber harvest. Thinning of the sites can upgrade the study area diversity and Abella (2010) emphasized that thinned plot can increase diversity richness up to three times more than un-thinned plots.

5.2. Relationship of Viscum album in Pinus sylvestris with elevation and DBH Even in the small range of elevation (1420-1650m) there is a significant negative (r=-0.5135; p<0.00) correlation between the Viscum album and elevation. The incidence of high density Viscum album in Pinus sylvestris is high at lower elevation and low in higher elevation. A significant negative relationship (r=-0.7; p<0.01) of incidence of Viscum album in conifers (silver fir) with elevation (up to 600 m-very high incidence) was observed in Eastern Carpathians (Barbu 2012). Dobbertin et al. (2005a) also found that around 37% of his sampled Pinus sylvestris trees were affected by Viscum album up to an elevation range of 1550 m and absent above this range. However, we found a positive correlation and significant relationship (r=0.52; p<0.00) between the DBH and incidence of Viscum album in Pinus sylvestris. The higher the DBH the more susceptible to incidence of Viscum album and other diseases. Barbu (2010) and Noetzli et al. (2003) also indicate that Pinus sylvestris trees more than 50 years old are more susceptible to incidence of Viscum album. Similarly, the mistletoe incidence on silver fir showed that is higher in trees older than 120 years (Barbu, 2010). Elevation and DBH are successful explanatory predictors in the linear multiple and logistic models. The relation of the models are weak although significant (p<0.00) at 95% confidence interval. However, species distributions and its responses depend on several environmental variables (Austin, 2007). The models may be improved by testing other variables such as a broader elevation range, tree density, tree height, slope, aspect, temperature and rainfall. Furthermore, other models like Maxent, (van Gils et al., 2012), neural network and Bayesian models (Guisan & Zimmermann, 2000; Joshi et al., 2006) can be compared with multiple linear and logistic models.


34

5.3. Analysis of spectral reflectance and vegetation indices In the presence of Viscum album, Pinus sylvestris has a low reflectance in all spectral bands (Visible and NIR bands) whereas high reflectance is shown by Pinus sylvestris in absence of Viscum album in the NIR band. Photosynthetically active vegetation has a higher reflectance in the NIR band than stressed and photosynthetically inactive vegetation (Beeri et al., 2007; Carter & Knapp, 2001) thus, the Pinus sylvestris in presence of Viscum album may be under stress and photosynthetically less active compared to Pinus sylvestris in the absence of Viscum album. A research by Eitel et al. (2011) has shown that the red edge band has the capability to detect the stress in plants due to changes in Chlab; a significant result in early stress detection in a conifer forest was found. According to Carter and Knapp (2001), transmittance, reflectance and absorptance of the stressed and healthy vegetation can be better distinguished in NIR. Similarly, in this research the Pinus sylvestris with Viscum album resulted in a decrease in absorption in the red band and a shift to the red edge and a larger difference in band NIR1 and NIR2. NDVI 85 has the potential to differentiate the presence and absence of Viscum album in Pinus sylvestris. Ismail et al. (2008) and Ismail et al. (2006) also revealed that NDVI shows the better performance compared to other vegetation indices in detecting the Pinus patula tree infestation by Sirex noctilio. Similarly, the previous research done by Bhattarai et al. (2011) has also obtained a significant result (p-value =-0.09) to discriminate Sirex woodwasp infested Pinus sylvestris in WV-2 by using NDVI 85. NDVI 85 value of Pinus sylvestris trees was lower in the presence of Viscum album which indicates a reduced in the total green biomass (chlorophyll) or leaf area index (LAI) of the Pinus trees. According to Adams et al. (1999) there is less absorption by chlorophyll and less reflectance in the near infrared region in the stressed vegetation.

5.4. Image classification and accuracy assessment

5.4.1. Pinus sylvestris distribution map Our overall classification accuracy (96%) and Kappa value (0.84) of our research showed the high standard (acceptable level >85%) of classification accuracy for optical data (Ismail & Jusoff, 2008). Our classification accuracy is high compared with the overall accuracy of 86% and overall kappa value of 0.77 for classifying three classes (coniferous, broadleaf and non-forest) by applying a support vector machine classifier in multispectral imagery (Dalponte et al., 2012). Similarly, Dalponte et al. (2012) has been able to increase the overall accuracy level to 91.5% and overall kappa (0.86) for the same three classes (coniferous, broadleaf and non-forest) by using spectral bands and LiDAR extracted tree height features. Likewise, Holmgren et al. (2008) has achieved an overall classification accuracy of 91% for Norway spruce (Picea abies), Scots pine (Pinus sylvestris) and deciduous trees using multispectral aerial imagery. The reasons behind the high classification accuracy in our research compared to other researches could be the high number of ground reference data (>250) (Ismail & Jusoff, 2008), low number of classes (Immitzer et al., 2012a), low diversity forest with distinct even-aged trees (conifers and broadleaf) (Digital Globe, 2009) and additional spectral bands (red edge, NIR2 ) in WV-2 image. Tree species classification by Immitzer et al. (2012a) using WV-2 imagery has improved the classification accuracy from 84 to 95% when he reduce the classes from 10 to 4 (namely; Picea abies, Pinus sylvestris, Fagus sylvatica, Querus robur). This confirmed that the classification accuracy increases as the number of classes is lower. Similarly, Ismail and Jusoff (2008) has improved the classification accuracy from 83 to 89% when he increases the reference data. Furthermore, red edge, NIR1 and NIR2 of WV-2 image plays a key role in class separability (Immitzer et al., 2012a). Additional spectral band of WV-2 identified coniferous and broadleaf trees with 99% accuracy (Immitzer et al., 2012b). Similarly, Pu and Landry (2012) improved the overall classification accuracy of mapping


35

urban tree species on WV-2 compared to IKONOS due to additional bands especially yellow, red-edge and NIR2 of WV-2.

5.4.2. Detection of Viscum album in Pinus sylvestris in WorldView-2 and the Ortho image Overall classification accuracy for the WV-2 (86%) and Ortho image (77%) for detection of Viscum album in Pinus sylvestris is somewhat lower than the Meddens et al. (2011) findings. Meddens et al. (2011) applied MLC to detect the mountain pine beetle attack using digital aerial imagery with spatial resolutions of 30cm, 1.2m, 2.4m and 4.2m and obtained the highest overall accuracy (90%) and kappa (0.88) in 2.4m spatial resolution. Spruce et al. (2011) obtained the overall accuracy of 88% for detecting forest defoliation by gypsy moth using coarse resolution (250 m) MODIS NDVI time series. Kantola et al. (2010) has achieved a higher accuracy (88%) for classification of defoliated Pinus sylvestris trees caused by pine sawfly by combining high pulse density airborne laser scanning (ALS) data and aerial digital imagery. Similarly, Rencz and Nemeth (1985) also achieve a higher accuracy of more than 80% for detecting the Dendrococtonus ponderosae infestation in Pinus trees by using coarse resolution (30 m) Landsat MSS. A comparison of the 2m resolution WV-2 and 0.15m resolution Ortho photo classification outputs shows that the WV-2 imagery of lower spatial but higher spectral resolution gave 9% higher classification accuracy. This is interesting to note as it indicates that high spatial resolution does not necessarily result in better Viscum album detection. Many researchers (Bhattarai et al., 2011; Meddens et al., 2011; Quackenbush et al., 2000; Rencz & Nemeth, 1985; Wulder et al., 2006) shows a finer similar results of weak detection on finer spatial resolution imagery for detection of parasites. Spectral reflectance values convey information on biophysical, biochemical and physiological characteristics of vegetation features and therefore can be used to distinguish vegetation along the spectrum (citation). Since MLCs rely on the ‘spectral signature’ to distinguish between species, the higher the spectral resolution, the more vegetation biophysical, biochemical and physiological characteristics may be detected from spectral data. The WV-2 imageries have four extra spectral bands that support vegetation identification. These include the coastal band that supports vegetation identification based upon its chlorophyll and water penetration characteristics. The yellow band which enhances identification of "yellow-ness" characteristics of targets (Digital Globe, 2009)and the red edge band that aids in the analysis of vegetative condition (Mutanga & Skidmore, 2004). With these extra spectral bands, the spectral signatures of Pinus and Viscum album are clearer in the WV-2 than in the orthophoto hence explaining the higher classification accuracies. On the other hand, although Viscum album is easily visually distinguishable, in higher spatial resolution imagery, there were mixed pixels in the classification based on the orthophoto. Yu et al (2006) attribute the salt and pepper noise to sun view geometry or bi-directional reflectance effects (Immitzer et al., 2012a; Immitzer et al., 2012b). Many pixels represent a single object in fine resolution imagery. Conifer crowns have even more mixed pixel effects due to low sunlit portions of trees (Immitzer et al., 2012b). However, WV-2 and Ortho image were able to detect Viscum album in Pinus sylvestris with >75% accuracy. Our result may reflect the impact of Viscum album on the canopy of Pinus sylvestris (host) and not the spectral reflectance of Viscum album itself. In the ground survey the condition of the host could receive more attention.

5.5. Sources of error in image classification for detection of Viscum album in optical imagery Notwithstanding advancement in remote sensing and availability of high spatial and spectral resolution satellite imagery, there are still challenges and uncertainties in accurately detecting and mapping the Viscum album in the Pinus sylvestris forest. The major challenges and problem faced during this research are discussed below:


36

Positional error Some errors occurred while using Garmin GPS and iPAQ to navigate the sampling plot to identify tree locations on the ground. The accuracy of the GPS and iPAQ ranges from 3 to 10 meters. Errors were due to poor satellite visibility and weak signal reception by the GPS because of canopy obstructions and cloud cover. For the mapping and detection of Viscum album at tree level, a high precision spatial accuracy is required. This condition was not fully fulfilled during this survey. Our DGPS was unable to connect a local base station. To get a cm level accuracy a DGPS needs to install. It will take at least 3-4 hours to establish a base station. This is time consuming, expensive, and not suitable in the study area due to rugged terrain and dense forest. However, to minimize the spatial error, LiDAR based canopy height model and printed 15cm spatial resolution Ortho map were used in the field to identify every sampled tree. Advantage of this high spatial resolution Ortho map was to identify and cross check the sample locations periphery.

Sampling error due to dense canopy cover Field sampling data plays a vital role in our research. However, some crucial errors occurred while recording the presence and absence of Viscum album in Pinus sylvestris on the ground. Mostly Viscum album is in the tree crown of the Pinus sylvestris and hard to see from the ground due to the dense canopy cover. Some of the sampling trees were recorded as without Viscum album in the field, while observation from the elevated areas and the Ortho photo (15cm) often revealed the presence of Viscum album. To minimize this sampling error the suspicious samples were removed from the analysis. Bird’s eye view is required to establish the presence of Viscum album in Pinus sylvestris. Furthermore, this research also indicates the advantages of aerial survey over ground survey for detection of Viscum album in Pinus sylvestris.


37

6. CONCLUSIONS AND RECOMMENDATIONS

6.1. Conclusions The research demonstrates the potential of high resolution optical satellite imagery (WV-2) and aerial ortho imagery for detecting Viscum album in Pinus sylvestris. Pinus sylvestris was classified with a high accuracy (96%). We are able to detect and map the Viscum album in Pinus sylvestris of Bois noir, Barcelonnette for the first time. The following conclusions were drawn out from the research findings:

Is Pinus sylvestris with and without Viscum album is spectrally distinct? From the analysis of spectral characteristics of Pinus sylvestris in WV-2 image, it was revealed that there is a distinct spectral reflectance of Pinus sylvestris in the presence and absence of Viscum album. The Red edge, NIR1 and NIR2 bands of WV-2 show better separability.

Can vegetation indices (NDVI-with different band combination) derived from multispectral imagery (WV-2) differentiate the presence and absence of Viscum album in Pinus sylvestris? Among three vegetation indices, NDVI 85 shows the highest performance in distinguishing the Viscum album in Pinus sylvestris.

Is there any significant relationship in incidence of Viscum album in Pinus sylvestris with elevation and DBH? There is a significant negative correlation (r=-0.5135; p<0.00) and significant positive correlation (r=0.52; p<0.00) with elevation and DBH respectively for the incidence of Viscum album in Pinus sylvestris. Furthermore, the linear and logistic regression model show weak predictions although significant (p<0.00) at 95% confidence interval by including both elevation and DBH.

How accurately can the incidence of Viscum album in Pinus sylvestris forest be detected by the use of maximum likelihood classification (pixel based classification) on WV-2 satellite and airborne Ortho imagery? An overall classification accuracy of 86% and a kappa of 0.52 was achieved on WV-2 for the detection of Viscum album, whereas in the airborne ortho image an accuracy (77%) and kappa (0.36). The user accuracy for both images was around 72% for the detection of Viscum album in Pinus sylvestris. Finally, the research indicated that the Viscum album can be detected in Pinus sylvestris by using remotely sensed very high resolution optical imagery.

6.2. Recommendations The issue of high incidence of Viscum album in the tree species is regarded as an alarming concern for the continuation of Pinus sylvestris forest. Pinus sylvestris one of the keystone tree species has been dramatically affected by the presence of Viscum album in the study area. Pinus sylvestris forests are threatened throughout the Alps. Zuber (2004b) cited that Dicrano Pinion, Erica-Pinion and Ononido-Pinion are the main Pinus species communities where Viscum album can grow. My recommendations are as follows:

There are no sign of thinning and management of forest over a long period of plantations which need to harvest and thinned. Thinning of the dense Pinus sylvestris can help to maintain the diversity of the forest. Thinning also enhance the understory vegetation as well (Boakye et al., 2012).


38

Substitution of mono-culture plantation of Pinus sylvestris by poly-culture plantation of other tree species which is less susceptible to incidence of Viscum album such as Pinus uncinata, Larix decidua. Since these species are growing at the same habitat in the study area in absence of Viscum album.

Positional error and correct observations in the ground need to be addressed.

The predictive model was developed on the basis of elevation and DBH only. Additional

explanatory variables such as slope, aspect, tree height, and climate can be used to develop models. Other models like Maxent, Neural Network and Bayesian models maybe tested as well.

Dobbertin et al. (2005a) reflects the incidence of Viscum album due to climate change (global

warming). Is really climate change affecting the impact of Viscum album on Pinus sylvestris? Climatic variables need to study for better understand the relationship of incidence of Viscum album on Pinus species. Further research on ecology of Viscum album and Pinus sylvestris is recommended to understand the phenomenon of incidence of Viscum album on Pinus species.

Spraying biological herbicides in the affected trees may minimize the growth rate of Viscum album.

Wind and birds are the major vectors for the dispersal of Viscum album. Considering these (wind

and bird migration) factors might be useful for minimizing the growth of Viscum album in Pinus species.

Spatial distribution of Viscum album on host species is still mysterious, for example Pinus sylvestris is

highly affected by Viscum album in Bois noir whereas there is no single incidence of Viscum album in Pinus in the Netherlands. What cause the Viscum album to select its host in a selective geographical area? What are the environmental and ecological factors for the incidence of Viscum album? Further research can be done on these queries.


39

LIST OF REFERENCES Abella, S. (2010). Thinning Pine Plantations to Reestablish Oak Openings Species in Northwestern Ohio.

Environmental Management, 46(3), 391-403. Adams, M. L., Philpot, W. D., & Norvell, W. A. (1999). Yellowness index: An application of spectral second

derivatives to estimate chlorosis of leaves in stressed vegetation. International Journal of Remote Sensing, 20(18), 3663-3675.

Allouche, O., Tsoar, A., & Kadmon, R. (2006). Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology, 43(6), 1223-1232.

Amer, B., Juvik, O. J., Dupont, F., Francis, G. W., & Fossen, T. (2012). Novel aminoalkaloids from European mistletoe (Viscum album L.). Phytochemistry Letters, 5(3), 677-681.

Astola, H., Sirro, L., Hame, T., Molinier, M., & Ahola, J. (2006). Separation of Coniferous Species in Boreal Forest Using Spectral and Contextual Features from Ikonos Imagery. New York: IEEE.

Austin, M. (2007). Species distribution models and ecological theory: A critical assessment and some possible new approaches. Ecological Modelling, 200(1-2), 1-19.

Barbu, C. (2010). The incidence and distribution of white mistletoe (Viscum album ssp. abietis) on Silver fir (Abies alba Mill.) stands from Eastern Carpathians. Annals of Forest Research, 53(1), 27-36.

Barbu , C. O. (2012). Impact of White Mistletoe (Viscum album ssp abietis) Infection on Needles and Crown Morphology of Silver Fir (Abies alba Mill.). Notulae Botanicae Horti Agrobotanici Cluj-Napoca, 40(2), 152-158.

Barney, C. W., Hawksworth, F. G., & Geils, B. W. (1998). Hosts of Viscum album. European Journal of Forest Pathology, 28(3), 187-208.

Beeri, O., Phillips, R., Hendrickson, J., Frank, A. B., & Kronberg, S. (2007). Estimating forage quantity and quality using aerial hyperspectral imagery for northern mixed-grass prairie. Remote Sensing of Environment, 110(2), 216-225.

Bhattarai, N., Quackenbush, L. J., Calandra, L., Im, J., & Teale, S. (2011). Spectral analysis of Scotch pine infested by Sirex noctilio. Paper presented at the American Society for Photogrammetry and Remote Sensing 2011, Milwaukee, Wisconsin.

Blanco, E., Bonet, J. A., & Eizaguirre, M. (2009). Using Landsat satellite imagery to detect small-size forest stands of Pinus nigra Arn. and Pinus sylvestris L. affected by Scolytidae. Investigacion Agraria-Sistemas Y Recursos Forestales, 18(3), 264-275.

Boakye, E. A., van Gils, H., Osei, E. M., & Asare, V. N. A. (2012). Does forest restoration using taungya foster tree species diversity? The case of Afram Headwaters Forest Reserve in Ghana. African Journal of Ecology, 50(3), 319-325.

Breshears, D. D., Cobb, N. S., Rich, P. M., Price, K. P., Allen, C. D., Balice, R. G., . . . Meyer, C. W. (2005). Regional vegetation die-off in response to global-change-type drought. Proceedings of the National Academy of Sciences of the United States of America, 102(42), 15144-15148.

Buddenbaum, H., Schlerf, M., & Hill, J. (2005). Classification of coniferous tree species and age classes using hyperspectral data and geostatistical methods. International Journal of Remote Sensing, 26(24), 5453-5465.

Carter, & Knapp, A. K. (2001). Leaf optical properties in higher plants: Linking spectral characteristics to stress and chlorophyll concentration. American Journal of Botany, 88(4), 677-684.

Carter , M. R. (1987). Seedling growth and mineral nutrition of Scots pine under acidic to calcareous soil conditions. Soil Science, 144(3), 175-180.

Chen, Q., Laurin, G. V., Battles, J. J., & Saah, D. (2012). Integration of airborne lidar and vegetation types derived from aerial photography for mapping aboveground live biomass. Remote Sensing of Environment, 121, 108-117.

Clark-Tapia, R., Torres-Bautista, B., Alfonso-Corrado, C., Valdez-Hernandez, J. I., Gonzalez-Adame, G., Bretado-Velazquez, J., & Campos-Contreras, J. (2011). Analysis of the abundance and mistletoe infection in Sierra Fria, Aguascalientes, Mexico. Madera Y Bosques, 17(2), 19-33.

Congalton, R. G. (1991). A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment, 37(1), 35-46.


40

Coops, N. C., Johnson, M., Wulder, M. A., & White, J. C. (2006). Assessment of QuickBird high spatial resolution imagery to detect red attack damage due to mountain pine beetle infestation. Remote Sensing of Environment, 103(1), 67-80.

Coops , N. C., Stone, C., Culvenor, D. S., & Chisholm, L. (2004). Assessment of crown condition in eucalypt vegetation by remotely sensed optical indices. Journal of Environmental Quality, 33(3), 956-964.

Cregg, B. M., & Zhang, J. W. (2001). Physiology and morphology of Pinus sylvestris seedlings from diverse sources under cyclic drought stress. Forest Ecology and Management, 154(1–2), 131-139.

Dalponte, M., Bruzzone, L., & Gianelle, D. (2012). Tree species classification in the Southern Alps based on the fusion of very high geometrical resolution multispectral/hyperspectral images and LiDAR data. Remote Sensing of Environment, 123(0), 258-270.

Dawson, T. E., Ehleringer, J. R., & Marshall, J. D. (1990). Sex-ratio and reproductive variation in the mistletoe Phoradendron juniperinum (Viscaceae). American Journal of Botany, 77(5), 584-589.

Digital Globe. (2009). WHITE PAPER: The benefits of the 8 Spectral Bands of WorldView -2 Retrieved 8/8/2012, 2012, from http://worldview2.digitalglobe.com/docs/WorldView-2_8-Band_Applications_Whitepaper.pdf

Dobbertin, M., Hilker, N., Rebetez, M., Zimmermann, N. E., Wohlgemuth, T., & Rigling, A. (2005a). The upward shift in altitude of pine mistletoe (Viscum album ssp austriacum) in Switzerland - the result of climate warming? International Journal of Biometeorology, 50(1), 40-47.

Dobbertin, M., & Rigling, A. (2006). Pine mistletoe (Viscum album ssp austriacum) contributes to Scots pine (Pinus sylvestris) mortality in the Rhone valley of Switzerland. Forest Pathology, 36(5), 309-322.

Dottavio, C. L., & Williams, D. L. (1983). Satellite Technology: An Improved Means for Monitoring Forest Insect Defoliation. Journal of Forestry, 81(1), 30-34.

Eitel, J. U. H., Vierling, L. A., Litvak, M. E., Long, D. S., Schulthess, U., Ager, A. A., . . . Stoscheck, L. (2011). Broadband, red-edge information from satellites improves early stress detection in a New Mexico conifer woodland. Remote Sensing of Environment, 115(12), 3640-3646.

EUFORGEN. (2009). Distribution map of Scots pine (Pinus sylvestris) Retrieved 2013/1/21, 2013, from http://www.euforgen.org/fileadmin/www.euforgen.org/Documents/Maps/PDF/Pinus_sylvestris.pdf

Falkenström, H., & Ekstrand, S. (2002). Evaluation of IRS-1c LISS-3 satellite data for defoliation assessment on Norway spruce and Scots pine. Remote Sensing of Environment, 82(2–3), 208-223.

Fauvart, N., Ali, A. A., Terral, J.-F., Roiron, P., Blarquez, O., & Carcaillet, C. (2012). Holocene upper tree-limits of Pinus section sylvestris in the Western Alps as evidenced from travertine archives. Review of Palaeobotany and Palynology, 169(0), 96-102.

Glatzel, G., & Geils, B. W. (2009). Mistletoe ecophysiology: host-parasite interactions. Botany-Botanique, 87(1), 10-15.

Gómez-Aparicio, L., Zavala, M. A., Bonet, F. J., & Zamora, R. (2009). Are Pine Plantations Valid Tools for Restoring Mediterranean Forests? An Assessment along Abiotic and Biotic Gradients. Ecological Applications, 19(8), 2124-2141.

Guisan, A., & Zimmermann, N. E. (2000). Predictive habitat distribution models in ecology. Ecological Modelling, 135(2–3), 147-186.

Günter, S., Weber, M., Erreis, R., & Aguirre, N. (2007). Influence of distance to forest edges on natural regeneration of abandoned pastures: a case study in the tropical mountain rain forest of Southern Ecuador. European Journal of Forest Research, 126(1), 67-75.

Haara, A., & Haarala, M. (2002). Tree Species Classification using Semi-automatic Delineation of Trees on Aerial Images. Scandinavian Journal of Forest Research, 17(6), 556-565.

Hartley, M. J. (2002). Rationale and methods for conserving biodiversity in plantation forests. Forest Ecology and Management, 155(1–3), 81-95.

Hawksworth, F. G., & Scharpf, R. F. (1986). Spread of European mistletoe (Viscum album) in California, USA European Journal of Forest Pathology, 16(1), 1-5.

Hicke, J. A., & Logan, J. (2009). Mapping whitebark pine mortality caused by a mountain pine beetle outbreak with high spatial resolution satellite imagery. International Journal of Remote Sensing, 30(17), 4427-4441.

Holmgren, J., & Persson, A. (2004). Identifying species of individual trees using airborne laser scanner. Remote Sensing of Environment, 90(4), 415-423.


41

Holmgren, J., Persson, Å., & Söderman, U. (2008). Species identification of individual trees by combining high resolution LiDAR data with multi-spectral images. International Journal of Remote Sensing, 29(5), 1537-1552.

Husch, B., Beers, T. W., & Kershaw, J. A. (2003). Forest Mensuration (Fourth Edition ed.). New Jersey: John Wiley & Sons.

Hussin, Y. A., Kimani, J., Lubczyski, M., Chavarro, D., & Obakeng, O. (2006). High resolution, remote sensing and object-oriented classification of Savannah vegetation for mapping transpiration. Paper presented at the 6th International conference on earth observation & geoinformation scieces in support of Africa's development, Cairo, Egypt.

Idzojtic, M., Pernar, R., Glavas, M., Zebec, M., & Diminic, D. (2008). The incidence of mistletoe (Viscum album ssp abietis) on silver fir (Abies alba) in Croatia. Biologia, 63(1), 81-85.

Immitzer, M., Atzberger, C., & Koukal, T. (2012a). Suitability of WorldView-2 data for tree species classification with special emphasis on the four new spectral bands. Photogrammetrie Fernerkundung Geoinformation(5), 573-588.

Immitzer, M., Atzberger, C., & Koukal, T. (2012b). Tree Species Classification with Random Forest Using Very High Spatial Resolution 8-Band WorldView-2 Satellite Data. Remote Sensing, 4(9), 2661-2693.

Ismail, M. H., & Jusoff, K. (2008). Satellite data classification accuracy assessment based from reference dataset. International Journal of Computer and Information Science and Engineering, 96-102.

Ismail, R., Mutanga, O., & Biob, U. (2006). The use of high resolution airborne imagery for the detection of forest canopy damage by Sirex noctilio. Paper presented at the Proceedings of the International Precision Forestry Symposium, Stellenbosch University, Stellenbosch.

Ismail, R., Mutanga, O., & Bob, U. (2007). Forest health and vitality: the detection and monitoring of Pinus patula trees infected by Sirex noctilio using digital multispectral imagery. Southern Hemisphere Forestry Journal, 69(1), 39-47.

Ismail, R., Mutanga, O., Kumar, L., & Bob, U. (2008). Determining the optimal spatial resolution of remotely sensed data for the detection of Sirex noctilio infestations in pine plantations in KwaZulu-Natal, South Africa. South African Geographical Journal, 90(1), 22-31.

IUCN. (2012). Red List of Threatened Species Retrieved 8/5/2012, 2012, from http://www.iucnredlist.org/details/42418/0

Jackson, R. D., & Huete, A. R. (1991). Interpreting vegetation indices. Preventive Veterinary Medicine, 11(3–4), 185-200.

Ji, L., & Peters, A. J. (2007). Performance evaluation of spectral vegetation indices using a statistical sensitivity function. Remote Sensing of Environment, 106(1), 59-65.

Jones, E. W. (1945). The structure and reproduction of the virgin forest of the North temperate zone. New Phytologist, 44(2), 130-148.

Joshi, C., De Leeuw, J., van Andel, J., Skidmore, A. K., Lekhak, H. D., van Duren, I. C., & Norbu, N. (2006). Indirect remote sensing of a cryptic forest understorey invasive species. Forest Ecology and Management, 225(1–3), 245-256.

Kantola, T., Vastaranta, M., Yu, X. W., Lyytikainen-Saarenmaa, P., Holopainen, M., Talvitie, M., . . . Hyyppa, J. (2010). Classification of Defoliated Trees Using Tree-Level Airborne Laser Scanning Data Combined with Aerial Images. Remote Sensing, 2(12), 2665-2679.

Kappes, M. S., Malet, J. P., Remaitre, A., Horton, P., Jaboyedoff, M., & Bell, R. (2011). Assessment of debris-flow susceptibility at medium-scale in the Barcelonnette Basin, France. Natural Hazards and Earth System Sciences, 11(2), 627-641.

Kavzoglu, T., & Reis, S. (2008). Performance analysis of maximum likelihood and artificial neural network classifiers for training sets with mixed pixels. GIScience and Remote Sensing, 45(3), 330-342.

Kharuk, V. I., Ranson, K. J., Kozuhovskaya, A. G., Kondakov, Y. P., & Pestunov, I. A. (2004). NOAA/AVHRR satellite detection of Siberian silkmoth outbreaks in eastern Siberia. International Journal of Remote Sensing, 25(24), 5543-5555.

Kint, V. (2005). Structural development in ageing temperate Scots pine stands. Forest Ecology and Management, 214(1-3), 237-250.

Kumar, V. (2012). Forest inventory parameters and carbon mapping from airborne LIDAR. University of Twente Faculty of Geo-Information and Earth Observation (ITC), Enschede. Retrieved from http://www.itc.nl/library/papers_2012/msc/nrm/vinodkumar.pdf


42

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159-174.

Leckie, Donald, G., Jay, C., Gougeon, F. A., Sturrock, R. N., & Paradine, D. (2004). Detection and assessment of trees with Phellinus weirii (laminated root rot) using high resolution multi-spectral imagery. International Journal of Remote Sensing, 25(4), 793-818.

Leckie, D. G., Gougeon, F. A., Walsworth, N., & Paradine, D. (2003). Stand delineation and composition estimation using semi-automated individual tree crown analysis. Remote Sensing of Environment, 85(3), 355-369.

Lilja, S., & Kuuluvainen, T. (2005). Structure of old Pinus sylvestris dominated forest stands along a geographic and human impact gradient in mid-boreal Fennoscandia. Silva Fennica, 39(3), 407-428.

Liu, X., Skidmore, A. K., & van Oosten, H. H. (2000). Discrimination ability of neural network and maximum likelihood classifiers. In: ISPRS 2000 congress : geoinformation for all : Amsterdam, the Netherlands, 16-23 July, 2000. pp. 782-789.

Lust, N., Muys, B., & Nachtergale, L. (1998). Increase of biodiversity in homogeneous Scots pine stands by an ecologically diversified management. Biodiversity and Conservation, 7(2), 249-260.

Magurran, A. E. (1988). Ecological diversity and its measurement. London, Great Britain: Cambridge University Press.

Maquaire, O., Malet, J. P., Remaı̂tre, A., Locat, J., Klotz, S., & Guillon, J. (2003). Instability conditions of marly hillslopes: towards landsliding or gullying? The case of the Barcelonnette Basin, South East France. Engineering Geology, 70(1–2), 109-130.

Margalef, R. (1958). Temporal succession and spatial heterogeneity in phytoplankton. In: Perspectives in Marie biology, . Berkeley: University California Press.

Meddens, A. J. H., Hicke, J. A., & Vierling, L. A. (2011). Evaluating the potential of multispectral imagery to map multiple stages of tree mortality. Remote Sensing of Environment, 115(7), 1632-1642. doi: 10.1016/j.rse.2011.02.018

Mutanga, O., & Skidmore, A. K. (2004). Narrow band vegetation indices overcome the saturation problem in biomass estimation. International Journal of Remote Sensing, 25(19), 3999-4014.

Naydenova, V., & Jelev, G. (2009). Forest Dynamics Study Using Aerial Photos and Satellite Images with Very High Spatial Resolution. New York: IEEE.

Noetzli, K. P., Müller, B., & Sieber, T. N. (2003). Impact of population dynamics of white mistletoe (Viscum album ssp. abietis) on European silver fir (Abies alba). Annals of Forest Science, 60(8), 773-779.

NRCS. (2012). Plants Profile, . Natural Resources Conservation Service, United States Department of Agriculture, Retrieved 8/5/2012 http://plants.usda.gov/java/profile?symbol=VIAL2

Oberhuber, W. (2001). The role of climate in the mortality of Scots pine (Pinus sylvestris L.) exposed to soil dryness. Dendrochronologia, 19(1), 45-55.

Oker-Blom, P., Kotisaari, A., Kellomaki, S., Ross, J., & Smolander, H. (1986). Crown Projection Area of Young Pinus sylvestris: a Model and Its Test. Scandinavian Journal of Forest Research, 1(1-4), 67-74.

Padwick, C., Deskevich, M., Pacifici, F., & Smallwood, S. (2010). WorldView-2 Pan Sharpening. Paper presented at the ASPRS 2010 Annual Conference, San Diego, California. http://www.digitalglobe.com/downloads/WorldView-2_Pan-Sharpening.pdf

Peršoh, D., Melcher, M., Flessa, F., & Rambold, G. (2010). First fungal community analyses of endophytic ascomycetes associated with Viscum album ssp. austriacum and its host Pinus sylvestris. Fungal Biology, 114(7), 585-596.

Pourababaei, H., Asgari, F., Reif, A., & Abedi, R. (2012). Effect of plantations on plant species diversity in the Darabkola, Mazandaran Province, North of Iran Bidiversitas, 13(2), 72-78.

Pu, R., & Landry, S. (2012). A comparative analysis of high spatial resolution IKONOS and WorldView-2 imagery for mapping urban tree species. Remote Sensing of Environment, 124, 516-533.

Quackenbush, L. J., Hopkins, P. F., & Kinn, G. J. (2000). Developing forestry products from high resolution digital aerial imagery. Photogrammetric Engineering and Remote Sensing, 66(11), 1337-1346.

Razak, K. A., Bucksch, A., Damen, M., van Westen, C., Straatsma, M., & de Jong, S. (2011a). Characterizing tree growth anomaly induced by landslides using LiDAR. Paper presented at the Second World Landslide Forum, Rome.


43

Razak, K. A., Straatsma, M. W., van Westen, C. J., Malet, J. P., & de Jong, S. M. (2011b). Airborne laser scanning of forested landslides characterization: Terrain model quality and visualization. Geomorphology, 126(1-2), 186-200.

Rencz, A. N., & Nemeth, J. (1985). Detection of mountain pine beetle infestation using Landsat MSS and simulated thematic mapper data. Canadian Journal of Remote Sensing, 11(1), 50-58.

Richardson, D. M. (1998). Ecology and Biogeography of Pinus. (pp. 546). Cambridge: Cambridge University Press.

Rigling, A., Eilmann, B., Koechli, R., & Dobbertin, M. (2010). Mistletoe-induced crown degradation in Scots pine in a xeric environment. Tree Physiology, 30(7), 845-852.

Roche, J., Mitchell, F., & Waldren, S. (2009). Plant community ecology of <i>Pinus sylvestris , an extirpated species reintroduced to Ireland. Biodiversity and Conservation, 18(8), 2185-2203.

Rouse, J. W., Haas, R. H., Schell, J. A., Deering, D. W., & and Harlan, J. C. (1974). Monitoring the Vernal Advancement of Retrogradation of Natural Vegetation: NASA/GSFC, TypeIII, Final Report, Greenbelt, MD.

Saez, J. L., Astrade, L., Corona, C., Stoffel, M., Berger, F., Jancke, O., & Schoeneich, F. The Forest: An efficient spatio-temporal bioindicator of landslide activities.

Saez, J. L., Corona, C., Stoffel, M., Astrade, L., Berger, F., & Malet, J. P. (2012). Dendrogeomorphic reconstruction of past landslide reactivation with seasonal precision: the Bois Noir landslide, southeast French Alps. Landslides, 9(2), 189-203.

Sanguesa-Barreda, G., Linares, J. C., & Camarero, J. J. (2012). Mistletoe effects on Scots pine decline following drought events: insights from within-tree spatial patterns, growth and carbohydrates. Tree Physiology, 32(5), 585-598.

Shannon, C. E., & Weaver, W. (1962). The mathematical theory of communication (Vol. 19). Urbana, USA: Universitry of Illinois Press.

Sim, J., & Wright, C. C. (2005). The kappa statistic in reliability studies: Use, interpretation, and sample size requirements. Physical Therapy, 85(3), 257-268.

Spruce, J. P., Sader, S., Ryan, R. E., Smoot, J., Kuper, P., Ross, K., . . . Hargrove, W. (2011). Assessment of MODIS NDVI time series data products for detecting forest defoliation by gypsy moth outbreaks. Remote Sensing of Environment, 115(2), 427-437.

Stone, C., & Coops, N. C. (2004). Assessment and monitoring of damage from insects in Australian eucalypt forests and commercial plantations. Australian Journal of Entomology, 43, 283-292.

Summers , R. W., Mavor , R. A., MacLennan , A. M., & Rebecca , G. W. (1999). The structure of ancient native pinewoods and other woodlands in the Highlands of Scotland. Forest Ecology and Management, 119(1–3), 231-245.

Thiery, Y., Malet, J. P., Sterlacchini, S., Puissant, A., & Maquaire, O. (2007). Landslide susceptibility assessment by bivariate methods at large scales: Application to a complex mountainous environment. Geomorphology, 92(1–2), 38-59.

Trees for Life. (2012). Species Profile, Scots Pine. Retrieved 8/5/2012 http://www.treesforlife.org.uk/tfl.scpine.html

Tsopelas, P., Angelopoulos, A., Economou, A., & Soulioti, N. (2004). Mistletoe (Viscum album) in the fir forest of Mount Parnis, Greece. Forest Ecology and Management, 202(1–3), 59-65.

Vacchiano, G., Motta, R., Long, J. N., & Shaw, J. D. (2008). A density management diagram for Scots pine (Pinus sylvestris L.): A tool for assessing the forest's protective effect. Forest Ecology and Management, 255(7), 2542-2554.

Vallauri, D. R., Aronson, J., & Barbero, M. (2002). An analysis of forest restoration 120 years after reforestation on badlands in the Southwestern Alps. Restoration Ecology, 10(1), 16-26.

van Gils, H., Conti, F., Ciaschetti, G., & Westinga, E. (2012). Fine resolution distribution modelling of endemics in Majella National Park, Central Italy. Plant Biosystems - An International Journal Dealing with all Aspects of Plant Biology, 146(sup1), 276-287.

van Gils, H., Delfino, J., Rugege, D., & Janssen, L. (2004). Efficacy of Chromolaena odorata control in a South African conservation forest. South African Journal of Science, 100(5-6), 251-253.

van Gils, H., Mwanangi, M., & Rugege, D. (2006). Invasion of an alien shrub across four land management regimes, west of St Lucia, South Africa. South African Journal of Science, 102(1-2), 9-12.


44

Varga, I., Taller, J., Baltazar, T., Hyvonen, J., & Poczai, P. (2012). Leaf-spot disease on European mistletoe (Viscum album) caused by Phaeobotryosphaeria visci: a potential candidate for biological control. Biotechnology Letters, 34(6), 1059-1065.

Vertui, F., & Tagliaferro, F. (1998). Scots pine (Pinus sylvestris L.) die-back by unknown causes in the Aosta Valley, Italy. Chemosphere, 36(4–5), 1061-1065.

Viera, A. J., & Garrett, J. M. (2005). Understanding interobserver agreement: the kappa statistic. Family Medicine, 37(5), 360-363.

Wallden, B. (1961). Misteln vid dess nordgräns (pp. 427 549). Wang, Adiku, S., Tenhunen, J., & Granier, A. (2005). On the relationship of NDVI with leaf area index in a

deciduous forest site. Remote Sensing of Environment, 94(2), 244-255. Warrens, M. J. (2011). Cohen’s kappa is a weighted average. Statistical Methodology, 8(6), 473-484. Watson, D. M. (2001). Mistletoe - A keystone resource in forests and woodlands worldwide. Annual Review of

Ecology and Systematics, 32, 219-249. Watt, M. S., & Kirschbaum, M. U. F. (2011). Moving beyond simple linear allometric relationships between

tree height and diameter. Ecological Modelling, 222(23–24), 3910-3916. Wulder, M. (1998). Optical remote-sensing techniques for the assessment of forest inventory and biophysical

parameters. Progress in Physical Geography, 22(4), 449-476. Wulder, M. A., Dymond, C. C., White, J. C., Leckie, D. G., & Carroll, A. L. (2006). Surveying mountain pine

beetle damage of forests: A review of remote sensing opportunities. Forest Ecology and Management, 221(1–3), 27-41.

Wulder, M. A., White, J. C., Coggins, S., Ortlepp, S. M., Coops, N. C., Heath, J., & Mora, B. (2012). Digital high spatial resolution aerial imagery to support forest health monitoring: the mountain pine beetle context. Journal of Applied Remote Sensing, 6(1), 062527-062521.

Xenakis, G., Ray, D., & Mencuccini, M. (2012). Effects of climate and site characteristics on Scots pine growth. European Journal of Forest Research, 131(2), 427-439.

Xie, Y. C., Sha, Z. Y., & Yu, M. (2008). Remote sensing imagery in vegetation mapping: a review. Journal of Plant Ecology, 1(1), 9-23.

Yang, C., Everitt, J. H., & Murden, D. (2011). Evaluating high resolution SPOT 5 satellite imagery for crop identification. Computers and Electronics in Agriculture, 75(2), 347-354.

Zhang, Y., Chen, H. Y. H., & Reich, P. B. (2012). Forest productivity increases with evenness, species richness and trait variation: a global meta-analysis. Journal of Ecology, 100(3), 742-749.

Zobel, D. B., Jha, P. K., Behan , M. J., & Yadav, U. K. (1985). A practical manuel for ecology. Kathmandu, Nepal: Ratna Book Distributors.

Zuber, D. (2004a). Biological flora of central Europe: Viscam album L. [Review]. Flora, 199(3), 181-203. Zuber, D. (2004b). Biological flora of Central Europe: Viscum album L. Flora - Morphology, Distribution,

Functional Ecology of Plants, 199(3), 181-203. Zuber, D. (2008). Biology and Evolution of the European Mistletoe (Viscum Album). Doctor of Sciences, ETH

Zurich. Zweifel, R., Bangerter, S., Rigling, A., & Sterck, F. J. (2012). Pine and mistletoes: how to live with a leak in the

water flow and storage system? Journal of Experimental Botany, 63(7), 2565-2578.


45

LIST OF APPENDICES: Appendix 1. Taxonomy of Pinus sylvestris and Viscum album :

Scientific

Name

Common

Name Kingdom Phylum Class Order Family

Red List

Category &

Criteria

Pinus

sylvestris

Pinus

sylvestris Plantae Tracheophyta Coniferopsida Coniferales Pinaceae

Lower Risk/Least

Concern

Source: (IUCN, 2012)

Scientific Name Common Name Group Family Duration

Viscum album L. European mistletoe Dicot Viscaceae Perennial

Source: (NRCS, 2012)

Appendix 2. Life cycle of Viscum album

Source: (Nierhaus-Wunderwald & Lawrenz 1997, Zuber 2004)


46

Appendix 3. Location map of the study area


47

Appendix 4. Features of WV-2 image

Sensor Name WorldView-2 Launch Information 13th September 2010 Acquisition Time 10:40:30

Spectral Range (nm)

Panchromatic 450-800 Coastal Blue 400-450 Blue 450-510 Green 510-580 Yellow 585-625 Red 630-690 Red-edge 705-745 Near Infrared 1 770-895 Near Infrared 2 860-1040

Sensor Resolution (m) ( GSD=Ground Sampling Distance)

Panchromatic 0.46 Multispectral 1.84

Radiometric Resolution 16 bits per pixel Sun Position 48.1° and 161.7°

DETE

CTIO

N AN

D MA

PPIN

G OF

INCI

DENC

E OF

VIS

CUM

ALBU

M IN

PIN

US S

YLVE

STRI

S FO

REST

IN S

OUTH

ERN

FREN

CH A

LPE

USIN

G SA

TELL

ITE

AND

AIRB

ORNE

OPT

ICAL

IMAG

ERY

48

App

endi

x 5.

Fi

eld d

ata

shee

t

Field

Dat

a She

et f

or M

appi

ng an

d de

tect

ion

of th

e V

iscum

albu

m in

Pinu

s sylv

estris

in B

ois n

oir,

Barc

elonn

ette

, Fra

nce

Dat

e:

Sam

plin

g M

etho

d *

Fo

rest

Typ

e

Coor

dina

tes

X

: Sa

mpl

e Pl

ot ID

Y:

Crow

n Co

vera

ge

Eas

t W

est

Cent

re

Nor

th

Sout

h A

vera

ge

Slop

e (%

):

Asp

ect:

A

ltitu

de:

Cano

py D

ensit

y:

Ave

rage

: Re

fere

nce

Poin

ts:

Be

arin

g:

Dist

ance

: S. N

. Tr

ee

ID

Spec

ies

D B

H

(cm)

Heig

ht

(m)

Crow

n di

amet

er

(m)

Coor

dina

te

Visc

um

album

lev

el**

Fore

st

Type

Dist

ance

fro

m

cent

re

poin

t Be

arin

g

Phot

o N

o.

Re

mar

ks

X

Y

1

321

49

1

2

3

21

491

3

321

49

1

4

3

21

491

5

321

49

1

6

3

21

491

7

321

49

1

8

3

21

491

9

321

49

1

10

321

49

1

N

ote:

Sa

mpl

ing

Type

**

1.

Tran

sect

Lin

e

2. C

ircul

ar P

lot

V

iscum

albu

m lev

el **

0=

No

Visc

um a

lbum

1=Lo

w V

iscum

albu

m

2=M

ediu

m V

iscum

albu

m

3=H

igh

Visc

um a

lbum


49

Appendix 6. Spatial location of sample plots

Plot ID Forest Type POINT_X POINT_Y Average Elevation P1 Pinus sylvestris 321557 4918741 1425 P2 Pinus sylvestris 321542 4918626 1457 P3 Pinus sylvestris 321604 4918617 1446 P4 Pinus sylvestris 321692 4918632 1437 P5 Pinus sylvestris 321426 4918417 1549 P6 Pinus sylvestris 321481 4918331 1517 P7 Mixed Coniferous 321324 4918270 1560 P8 Pinus sylvestris 321444 4918265 1537 P9 Pinus sylvestris 321550 4918268 1528 P10 Pinus sylvestris 321841 4918215 1482 P11 Pinus sylvestris 321401 4918212 1546 P12 Mixed Coniferous 321220 4918215 1568 P13 Pinus sylvestris 321191 4918141 1576 P14 Pinus sylvestris 321271 4918140 1568 P15 Pinus sylvestris 321380 4918135 1557 P16 Pinus sylvestris 321737 4918081 1513 P17 Pinus sylvestris 321514 4918098 1544 P18 Pinus sylvestris 321252 4918050 1579 P19 Pinus sylvestris 321435 4918039 1559 P20 Pinus sylvestris 321761 4917987 1530 P21 Pinus sylvestris 321678 4917975 1549 P22 Pinus sylvestris 321392 4917990 1571 P23 Pinus sylvestris 321328 4918008 1570 P24 Pinus sylvestris 321208 4917911 1598 P25 Pinus sylvestris 321256 4917879 1603 P26 Pinus sylvestris 321313 4917878 1594 P27 Pinus sylvestris 321410 4917893 1589 P28 Pinus sylvestris 321691 4917919 1559 P29 Pinus sylvestris 321513 4918463 1509 P30 Pinus sylvestris 321401 4918212 1546 P32 Pinus sylvestris 321382 4918365 1559


50

Appendix 7. LiDAR derived tree Canopy Height Model (CHM) of sample plot 30

(Source: Anahita & Collins)

Appendix 8. Descriptive statistics of the Bois noir

S.N. Species No. of specimen D RD F % RF % A RA %

1 Alnus viridis 2 1.29 0.16 3.23 1.30 2.00 2.93

2 Broadleaf deciduous

137 88.39 10.74 51.61 20.78 8.56 12.56

3 Fraxinus excelsior 37 23.87 2.90 19.35 7.79 6.17 9.05 4 Larix decidua 13 8.39 1.02 22.58 9.09 1.86 2.73 5 Picea abies 6 3.87 0.47 9.68 3.90 2.00 2.93 6 Pinus sylvestris 957 617.42 75.00 100.00 40.26 30.87 45.30 7 Pinus uncinata 103 66.45 8.07 29.03 11.69 11.44 16.79 8 Populas tremula 21 13.55 1.65 12.90 5.19 5.25 7.70 Total 1276 823.23 100.00 248.39 100.00 68.15 100.00 Where: S.N. = Serial Number D= density, RD= Relative Density, F= Frequency, RF= Relative Frequency A= Abundance , RA= Relative Abundance


51

Appendix 9. Histogram of Pinus sylvestris tree parameters


52

Appendix 10. Descriptive statistics of Pinus sylvestris tree parameter

Pinus sylvestris Variable n Min Max Mean Std. deviation Median Mode DBH (cm) 956 3 59 18.68 9.0 18 18 Height (m) 227 4 23 13.21 3.09 13 11 Elevation (m) 172 1423 1605 1550 34 1557 1568 Appendix 11. DBH class of Pinus sylvestris

Tree Category DBH Range Tree number % Sapling <10 cm 161 16.8 Pole 10-30 cm 681 71.2 Sawlog >30 cm 114 12 Total 956 100

Appendix 12. Descriptive statistics of presence and absence of Viscum album in Pinus sylvestris

Status of Viscum album in Pinus sylvestris

No. of plots in which Viscum album Occurred

D (tree/ha)

RD (%)

F %

RF %

A

RA %

BA/ha

RBA

Presence 17 40.6 6.6 54.8 35.4 3.7 11.4 3.3 16.1 Absence 31 577.4 93.4 100.0 64.6 28.9 88.6 17.5 83.9 Total 618.1 100.0 154.8 100.0 32.6 100.0 20.8 100.0 Where, D = Density RD= Relative Density F= Frequency; RF=Relative Frequency A= Abundance RA= Relative Abundance BA=Basal Area RBA=Relative Basal area Appendix 13. Kappa statistics of classified species

Species Kappa (K^) Pinus sylvestris 0.8291 Broadleaf 0.9361 Others 0.7939


53

Appendix 14. Field pictures

Pinus sylvestris tree Medium density Viscum album in Pinus

sylvestris tree

Heavy density Viscum album in Pinus Dying stage of Pinus sylvestris (Viscunm album

on crown of the tree)


54

Difficulties in observation of Viscum album

from ground due to high Dense canopy Cover (Photo Courtesy: Anahita Khosravipur)

High density of Viscum album in Pinus sylvestris forest

Setting up the DGPS

(Photo Courtesy: Collins Kukunda) Recording the co-ordinates of the sampling

point

Author navigating the plot with iPAQ

(Photo Courtesy: Collins Kukunda) Open field in the study area

(Photo Courtesy: Anahita Khosravipur)


55

Measuring DBH of tree Dense pole size Pinus sylvestris

Research Team